diff --git a/src/sentry/seer/agent/embed_widgets.generated.json b/src/sentry/seer/agent/embed_widgets.generated.json
index c4706cfec7a3..3e90f7da2d15 100644
--- a/src/sentry/seer/agent/embed_widgets.generated.json
+++ b/src/sentry/seer/agent/embed_widgets.generated.json
@@ -1113,6 +1113,136 @@
       }
     ]
   },
+  {
+    "name": "conversation",
+    "description": "The ONLY way to reference a single AI agent conversation (Explore > Agents). Use the `conversationId` exactly as the agents conversations API returns it. Include the API-provided `title` when available, and `start`/`end` (the conversation's own first and last span timestamps) so the embed can scope its query instead of scanning the default window. Inline: renders a compact link. Block: renders the conversation transcript with its LLM call, token, cost, and tool totals. Do not duplicate the messages or those totals as text. Never use a markdown link for conversation references.",
+    "level": ["inline", "block"],
+    "body": {
+      "$schema": "https://json-schema.org/draft/2020-12/schema",
+      "type": "object",
+      "properties": {
+        "id": {
+          "type": "string",
+          "minLength": 1
+        },
+        "title": {
+          "type": "string",
+          "minLength": 1
+        },
+        "projects": {
+          "type": "array",
+          "items": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "number"
+              }
+            ]
+          }
+        },
+        "start": {
+          "type": "string",
+          "format": "date-time",
+          "pattern": "^(?:(?:\\d\\d[2468][048]|\\d\\d[13579][26]|\\d\\d0[48]|[02468][048]00|[13579][26]00)-02-29|\\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\\d|30)|(?:02)-(?:0[1-9]|1\\d|2[0-8])))T(?:(?:[01]\\d|2[0-3]):[0-5]\\d(?::[0-5]\\d(?:\\.\\d+)?)?(?:Z|([+-](?:[01]\\d|2[0-3]):[0-5]\\d)))$"
+        },
+        "end": {
+          "type": "string",
+          "format": "date-time",
+          "pattern": "^(?:(?:\\d\\d[2468][048]|\\d\\d[13579][26]|\\d\\d0[48]|[02468][048]00|[13579][26]00)-02-29|\\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\\d|30)|(?:02)-(?:0[1-9]|1\\d|2[0-8])))T(?:(?:[01]\\d|2[0-3]):[0-5]\\d(?::[0-5]\\d(?:\\.\\d+)?)?(?:Z|([+-](?:[01]\\d|2[0-3]):[0-5]\\d)))$"
+        }
+      },
+      "required": ["id"],
+      "additionalProperties": false
+    },
+    "examples": [
+      {
+        "label": "Conversation",
+        "data": {
+          "id": "4821",
+          "title": "Refund request escalated to a human",
+          "start": "2026-08-25T16:37:12Z",
+          "end": "2026-08-25T16:39:02Z"
+        }
+      }
+    ],
+    "featureFlag": "organizations:gen-ai-conversations"
+  },
+  {
+    "name": "conversationsQuery",
+    "description": "Preview the AI agent conversations list (Explore > Agents) filtered by a search query. Use this when pointing the user at a SET of conversations — if you have a specific conversation ID, use the `conversation` embed instead. `query` uses span search syntax over gen_ai spans, e.g. \"gen_ai.request.model:gpt-4o\". Negation is not supported. Use `agents` to filter to specific agent names. Inline renders a link; block renders the first five matching conversations with their duration, message count, errors and cost.",
+    "level": ["inline", "block"],
+    "body": {
+      "$schema": "https://json-schema.org/draft/2020-12/schema",
+      "type": "object",
+      "properties": {
+        "projects": {
+          "description": "Project IDs. Omit for the \"My Projects\" selection.",
+          "type": "array",
+          "items": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "number"
+              }
+            ]
+          }
+        },
+        "environments": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "statsPeriod": {
+          "description": "Relative time range, e.g. \"24h\" or \"7d\". Mutually exclusive with start/end.",
+          "type": "string",
+          "pattern": "^\\d+[smhdw]$"
+        },
+        "start": {
+          "type": "string",
+          "format": "date-time",
+          "pattern": "^(?:(?:\\d\\d[2468][048]|\\d\\d[13579][26]|\\d\\d0[48]|[02468][048]00|[13579][26]00)-02-29|\\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\\d|30)|(?:02)-(?:0[1-9]|1\\d|2[0-8])))T(?:(?:[01]\\d|2[0-3]):[0-5]\\d(?::[0-5]\\d(?:\\.\\d+)?)?(?:Z|([+-](?:[01]\\d|2[0-3]):[0-5]\\d)))$"
+        },
+        "end": {
+          "type": "string",
+          "format": "date-time",
+          "pattern": "^(?:(?:\\d\\d[2468][048]|\\d\\d[13579][26]|\\d\\d0[48]|[02468][048]00|[13579][26]00)-02-29|\\d{4}-(?:(?:0[13578]|1[02])-(?:0[1-9]|[12]\\d|3[01])|(?:0[469]|11)-(?:0[1-9]|[12]\\d|30)|(?:02)-(?:0[1-9]|1\\d|2[0-8])))T(?:(?:[01]\\d|2[0-3]):[0-5]\\d(?::[0-5]\\d(?:\\.\\d+)?)?(?:Z|([+-](?:[01]\\d|2[0-3]):[0-5]\\d)))$"
+        },
+        "query": {
+          "default": "",
+          "type": "string"
+        },
+        "agents": {
+          "description": "Filter to these agent names, as reported by gen_ai.agent.name.",
+          "type": "array",
+          "items": {
+            "type": "string"
+          }
+        },
+        "title": {
+          "type": "string",
+          "minLength": 1
+        }
+      },
+      "required": ["query"],
+      "additionalProperties": false
+    },
+    "examples": [
+      {
+        "label": "Conversations with tool errors",
+        "data": {
+          "query": "gen_ai.tool.name:*",
+          "statsPeriod": "24h",
+          "title": "Conversations using tools"
+        }
+      }
+    ],
+    "featureFlag": "organizations:gen-ai-conversations"
+  },
   {
     "name": "replaysQuery",
     "description": "Preview the Session Replay list filtered by a search query. Use this when pointing the user at a SET of replays — if you have a specific replay ID, use the `replay` embed instead. `query` uses replay search syntax, e.g. \"user.email:user@example.com\". Inline renders a link; block renders the first five matching replays with their duration, error count and rage clicks.",
diff --git a/static/app/components/seer/markdown/__stories__/conversationEmbedStory.spec.tsx b/static/app/components/seer/markdown/__stories__/conversationEmbedStory.spec.tsx
new file mode 100644
index 000000000000..b7135e3176d8
--- /dev/null
+++ b/static/app/components/seer/markdown/__stories__/conversationEmbedStory.spec.tsx
@@ -0,0 +1,106 @@
+import {render, screen} from 'sentry-test/reactTestingLibrary';
+
+import type {Conversation} from 'sentry/views/explore/conversations/hooks/useConversations';
+
+import {ConversationEmbedStory} from './conversationEmbedStory';
+
+jest.mock('sentry/components/seer/markdown', () => ({
+  SeerMarkdown: ({raw}: {raw: string}) => <div aria-label="Rendered markdown">{raw}</div>,
+}));
+
+function createConversation(conversation: Partial<Conversation>): Conversation {
+  return {
+    conversationId: 'conv-1',
+    duration: 1000,
+    endTimestamp: Date.UTC(2026, 7, 25, 16, 39, 2),
+    errors: 0,
+    firstInput: 'Where is my refund?',
+    generationDuration: 800,
+    inputTokens: 100,
+    lastOutput: 'Escalating to a human.',
+    llmCalls: 3,
+    outputTokens: 50,
+    projectId: 11,
+    startTimestamp: Date.UTC(2026, 7, 25, 16, 37, 12),
+    title: null,
+    toolCalls: 1,
+    toolErrors: 0,
+    toolNames: ['lookup_order'],
+    totalCost: 0.02,
+    totalTokens: 150,
+    traceCount: 1,
+    traceIds: ['11111111111111111111111111111111'],
+    user: null,
+    ...conversation,
+  };
+}
+
+describe('ConversationEmbedStory', () => {
+  it('prefers a titled conversation and passes its ISO time bounds', async () => {
+    const untitled = createConversation({conversationId: 'untitled-conversation'});
+    const titled = createConversation({
+      conversationId: 'titled-conversation',
+      title: 'Refund request escalated to a human',
+    });
+    const conversationsRequest = MockApiClient.addMockResponse({
+      url: '/organizations/org-slug/agents/conversations/',
+      body: [untitled, titled],
+      match: [
+        MockApiClient.matchQuery({
+          project: [-1],
+          per_page: 25,
+          statsPeriod: '14d',
+        }),
+      ],
+    });
+
+    render(<ConversationEmbedStory />);
+
+    const renderedMarkdown = await screen.findByLabelText('Rendered markdown');
+    expect(renderedMarkdown).toHaveTextContent(titled.conversationId);
+    expect(renderedMarkdown).toHaveTextContent('Refund request escalated to a human');
+    expect(renderedMarkdown).toHaveTextContent(
+      new Date(titled.startTimestamp).toISOString()
+    );
+    expect(renderedMarkdown).toHaveTextContent(
+      new Date(titled.endTimestamp).toISOString()
+    );
+    expect(renderedMarkdown).not.toHaveTextContent(untitled.conversationId);
+    expect(conversationsRequest).toHaveBeenCalled();
+  });
+
+  it('falls back to the most recent conversation when none has a title', async () => {
+    const older = createConversation({
+      conversationId: 'older-conversation',
+      endTimestamp: Date.UTC(2026, 7, 24, 10, 0, 0),
+    });
+    const newer = createConversation({
+      conversationId: 'newer-conversation',
+      endTimestamp: Date.UTC(2026, 7, 26, 10, 0, 0),
+    });
+    // The endpoint orders by relevance, so the newest row is not necessarily first.
+    MockApiClient.addMockResponse({
+      url: '/organizations/org-slug/agents/conversations/',
+      body: [older, newer],
+    });
+
+    render(<ConversationEmbedStory />);
+
+    const renderedMarkdown = await screen.findByLabelText('Rendered markdown');
+    expect(renderedMarkdown).toHaveTextContent(newer.conversationId);
+    expect(renderedMarkdown).not.toHaveTextContent(older.conversationId);
+  });
+
+  it('renders a message when the organization has no conversations', async () => {
+    MockApiClient.addMockResponse({
+      url: '/organizations/org-slug/agents/conversations/',
+      body: [],
+    });
+
+    render(<ConversationEmbedStory />);
+
+    expect(
+      await screen.findByText('No conversation is available for this organization.')
+    ).toBeInTheDocument();
+  });
+});
diff --git a/static/app/components/seer/markdown/__stories__/conversationEmbedStory.tsx b/static/app/components/seer/markdown/__stories__/conversationEmbedStory.tsx
new file mode 100644
index 000000000000..5098837cad81
--- /dev/null
+++ b/static/app/components/seer/markdown/__stories__/conversationEmbedStory.tsx
@@ -0,0 +1,78 @@
+import {useQuery} from '@tanstack/react-query';
+
+import {Text} from '@sentry/scraps/text';
+
+import {LoadingIndicator} from 'sentry/components/loadingIndicator';
+import {ALL_ACCESS_PROJECTS} from 'sentry/components/pageFilters/constants';
+import {apiOptions} from 'sentry/utils/api/apiOptions';
+import {useOrganization} from 'sentry/utils/useOrganization';
+import type {Conversation} from 'sentry/views/explore/conversations/hooks/useConversations';
+
+import {EmbedStory, EmbedVariant} from './embedStory';
+
+/** Enough rows to find one with a title without paging the whole list. */
+const STORY_CONVERSATION_LIMIT = 25;
+
+/**
+ * The schema's `start`/`end` are ISO strings, but a conversation row carries
+ * epoch milliseconds.
+ */
+function toIsoTimestamp(timestamp: number | null | undefined): string | undefined {
+  return timestamp ? new Date(timestamp).toISOString() : undefined;
+}
+
+/**
+ * `useConversations` reads its filters from the router, which a story has no
+ * business driving, so the list is fetched directly -- the same call the
+ * `conversationsQuery` block makes.
+ */
+export function ConversationEmbedStory() {
+  const organization = useOrganization();
+  const {data, isError, isPending} = useQuery({
+    ...apiOptions.as<Conversation[]>()(
+      '/organizations/$organizationIdOrSlug/agents/conversations/',
+      {
+        path: {organizationIdOrSlug: organization.slug},
+        query: {
+          project: [ALL_ACCESS_PROJECTS],
+          per_page: STORY_CONVERSATION_LIMIT,
+          statsPeriod: '14d',
+        },
+        staleTime: 30_000,
+      }
+    ),
+    retry: false,
+  });
+
+  // The endpoint orders by relevance rather than recency, and a titled
+  // conversation shows off the embed better than an untitled one.
+  const conversations = (data ?? []).toSorted((a, b) => b.endTimestamp - a.endTimestamp);
+  const conversation = conversations.find(row => row.title) ?? conversations[0];
+
+  return (
+    <EmbedStory name="conversation">
+      {isPending ? (
+        <LoadingIndicator />
+      ) : isError ? (
+        <Text variant="muted">Unable to load a conversation example.</Text>
+      ) : conversation ? (
+        <EmbedVariant
+          name="conversation"
+          label="Conversation"
+          data={{
+            id: conversation.conversationId,
+            title: conversation.title ?? undefined,
+            projects:
+              conversation.projectId === null
+                ? undefined
+                : [String(conversation.projectId)],
+            start: toIsoTimestamp(conversation.startTimestamp),
+            end: toIsoTimestamp(conversation.endTimestamp),
+          }}
+        />
+      ) : (
+        <Text variant="muted">No conversation is available for this organization.</Text>
+      )}
+    </EmbedStory>
+  );
+}
diff --git a/static/app/components/seer/markdown/embeds/components/conversation/conversation.spec.tsx b/static/app/components/seer/markdown/embeds/components/conversation/conversation.spec.tsx
new file mode 100644
index 000000000000..bbd226047bb1
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversation/conversation.spec.tsx
@@ -0,0 +1,175 @@
+import {act, screen, waitFor} from 'sentry-test/reactTestingLibrary';
+
+import {PageFiltersStore} from 'sentry/components/pageFilters/store';
+import {
+  getEmbedLinkHref,
+  renderEmbed,
+} from 'sentry/components/seer/markdown/embeds/components/resourceEmbedTestUtils';
+
+const CONVERSATION_ID = 'conv-1';
+const DETAIL_URL = `/organizations/org-slug/agents/conversations/${CONVERSATION_ID}/`;
+
+/** Shaped as the conversation detail endpoint returns it: flat spans. */
+function spanFixture(overrides: Record<string, unknown>) {
+  return {
+    'gen_ai.conversation.id': CONVERSATION_ID,
+    parent_span: 'parent-1',
+    project: 'test-project',
+    'project.id': 1,
+    'span.status': 'ok',
+    trace: 'trace-1',
+    'gen_ai.operation.type': 'ai_client',
+    ...overrides,
+  };
+}
+
+function searchParams(href: string) {
+  return new URLSearchParams(href.split('?')[1] ?? '');
+}
+
+describe('conversation embed', () => {
+  beforeAll(async () => {
+    // The block is behind `lazy()`. Compiling its module graph on first render
+    // costs more than a `findBy*` will wait for, so pay it here instead of
+    // inside the first block assertion.
+    await import('./conversationBlock');
+  }, 60_000);
+
+  beforeEach(() => {
+    MockApiClient.clearMockResponses();
+    act(() => {
+      PageFiltersStore.reset();
+      PageFiltersStore.init();
+    });
+  });
+
+  it('links to the conversation detail view inline', () => {
+    const href = getEmbedLinkHref('conversation', 'Refund request escalated', {
+      id: CONVERSATION_ID,
+      title: 'Refund request escalated',
+      projects: ['1'],
+      start: '2026-08-25T16:37:12Z',
+      end: '2026-08-25T16:39:02Z',
+    });
+
+    expect(href.split('?')[0]).toBe(
+      `/organizations/org-slug/explore/agents/conversations/${CONVERSATION_ID}/`
+    );
+    const params = searchParams(href);
+    // The detail view scopes its span query to this window, so it is padded
+    // an hour either side of the conversation's own timestamps.
+    expect(params.get('start')).toBe('2026-08-25T15:37:12.000Z');
+    expect(params.get('end')).toBe('2026-08-25T17:39:02.000Z');
+    expect(params.getAll('project')).toEqual(['1']);
+    expect(params.get('referrer')).toBe('seer-conversation-embed');
+  });
+
+  it('falls back to the id when the tag carries no title', () => {
+    const href = getEmbedLinkHref('conversation', `Conversation ${CONVERSATION_ID}`, {
+      id: CONVERSATION_ID,
+    });
+
+    expect(href).toBe(
+      `/organizations/org-slug/explore/agents/conversations/${CONVERSATION_ID}/?referrer=seer-conversation-embed`
+    );
+  });
+
+  it('renders the aggregates but not the transcript', async () => {
+    const request = MockApiClient.addMockResponse({
+      url: DETAIL_URL,
+      body: {
+        conversationId: CONVERSATION_ID,
+        title: 'Out of memory investigation',
+        spans: [
+          spanFixture({
+            span_id: 'span-a',
+            'span.name': 'first turn',
+            'precise.start_ts': 1000,
+            'precise.finish_ts': 1000.5,
+            'gen_ai.request.messages': JSON.stringify([
+              {role: 'user', content: 'Why did the job fail?'},
+            ]),
+            'gen_ai.response.text': 'The worker ran out of memory.',
+            'gen_ai.usage.total_tokens': 1200,
+          }),
+        ],
+      },
+    });
+
+    renderEmbed({
+      name: 'conversation',
+      data: {
+        id: CONVERSATION_ID,
+        title: 'Stale title',
+        start: '2026-08-25T16:37:12Z',
+        end: '2026-08-25T16:39:02Z',
+      },
+    });
+
+    // The aggregates bar renders its labels while loading too, so the API
+    // title -- which only arrives with the response -- is the loaded signal.
+    // The API title also wins over whatever the model wrote into the tag.
+    expect(
+      await screen.findByRole('link', {name: 'Out of memory investigation'})
+    ).toBeInTheDocument();
+    expect(screen.queryByText('Stale title')).not.toBeInTheDocument();
+
+    expect(screen.getByText('LLM Calls')).toBeInTheDocument();
+    expect(screen.getByText('Errors')).toBeInTheDocument();
+    expect(screen.getByText('Tokens')).toBeInTheDocument();
+    expect(screen.getByText('Cost')).toBeInTheDocument();
+
+    // The embed renders inside an agent conversation, so it deliberately shows
+    // the totals only -- a nested transcript reads as part of the answer.
+    expect(screen.queryByText('The worker ran out of memory.')).not.toBeInTheDocument();
+    expect(screen.queryByText('Why did the job fail?')).not.toBeInTheDocument();
+
+    await waitFor(() => {
+      expect(request).toHaveBeenCalledWith(
+        DETAIL_URL,
+        expect.objectContaining({
+          query: expect.objectContaining({
+            // The tag's ISO timestamps scope the query instead of the
+            // default page-filter window.
+            start: '2026-08-25T15:37:12.000Z',
+            end: '2026-08-25T17:39:02.000Z',
+          }),
+        })
+      );
+    });
+  });
+
+  it("scopes the fetch to the tag's projects, not the host page's", async () => {
+    const request = MockApiClient.addMockResponse({
+      url: DETAIL_URL,
+      body: {conversationId: CONVERSATION_ID, title: 'Scoped', spans: []},
+    });
+
+    // The page the embed is rendered into is filtered to a different project.
+    act(() => {
+      PageFiltersStore.updateProjects([99], null);
+    });
+
+    renderEmbed({
+      name: 'conversation',
+      data: {id: CONVERSATION_ID, projects: ['7']},
+    });
+
+    await waitFor(() => {
+      expect(request).toHaveBeenCalledWith(
+        DETAIL_URL,
+        expect.objectContaining({
+          query: expect.objectContaining({project: [7]}),
+        })
+      );
+    });
+  });
+
+  it('shows an error when the conversation cannot be loaded', async () => {
+    MockApiClient.addMockResponse({url: DETAIL_URL, statusCode: 500, body: {}});
+
+    renderEmbed({name: 'conversation', data: {id: CONVERSATION_ID}});
+
+    expect(await screen.findByText('Unable to load conversation.')).toBeInTheDocument();
+  });
+});
diff --git a/static/app/components/seer/markdown/embeds/components/conversation/conversation.tsx b/static/app/components/seer/markdown/embeds/components/conversation/conversation.tsx
new file mode 100644
index 000000000000..569e64aa6499
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversation/conversation.tsx
@@ -0,0 +1,18 @@
+import {lazy} from 'react';
+
+import {LazyLoad} from 'sentry/components/lazyLoad';
+import {defineSeerEmbed} from 'sentry/components/seer/markdown/embeds/utils';
+
+import {ConversationLink} from './conversationLink';
+
+const LazyConversationBlock = lazy(() => import('./conversationBlock'));
+
+export const Conversation = defineSeerEmbed({
+  name: 'conversation',
+  render(props, level) {
+    if (level === 'block') {
+      return <LazyLoad LazyComponent={LazyConversationBlock} data={props} />;
+    }
+    return <ConversationLink data={props} />;
+  },
+});
diff --git a/static/app/components/seer/markdown/embeds/components/conversation/conversationBlock.tsx b/static/app/components/seer/markdown/embeds/components/conversation/conversationBlock.tsx
new file mode 100644
index 000000000000..2ba9c29648a6
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversation/conversationBlock.tsx
@@ -0,0 +1,64 @@
+import {Text} from '@sentry/scraps/text';
+
+import {QueryEmbedCard} from 'sentry/components/seer/markdown/embeds/components/queryEmbed/queryEmbedCard';
+import {t} from 'sentry/locale';
+import {ConversationAggregatesBar} from 'sentry/views/explore/conversations/components/conversationSummary';
+import {useConversation} from 'sentry/views/explore/conversations/hooks/useConversation';
+
+import {ConversationLink, type ConversationData} from './conversationLink';
+
+function toTimestampMs(isoTimestamp: string | undefined): number | undefined {
+  if (!isoTimestamp) {
+    return undefined;
+  }
+  const parsed = Date.parse(isoTimestamp);
+  return Number.isNaN(parsed) ? undefined : parsed;
+}
+
+/**
+ * Scopes the fetch to the conversation's own projects rather than the host
+ * page's filters, which have nothing to do with the conversation the tag names.
+ * The ids come from the model, so anything non-numeric is dropped instead of
+ * being sent as `NaN`.
+ */
+function toProjectIds(projects: ConversationData['projects']): number[] | undefined {
+  if (!projects?.length) {
+    return undefined;
+  }
+  const ids = projects.map(Number).filter(id => Number.isFinite(id));
+  return ids.length > 0 ? ids : undefined;
+}
+
+/**
+ * Deliberately shows the conversation's totals and not its transcript: the
+ * embed is itself rendered inside an agent conversation, so a nested transcript
+ * reads as part of the surrounding answer. The link goes to the full detail
+ * view for anyone who wants the messages.
+ */
+export default function ConversationBlock({data}: {data: ConversationData}) {
+  const {nodes, isLoading, error, title} = useConversation({
+    conversationId: data.id,
+    projects: toProjectIds(data.projects),
+    startTimestamp: toTimestampMs(data.start),
+    endTimestamp: toTimestampMs(data.end),
+  });
+
+  return (
+    <QueryEmbedCard
+      link={<ConversationLink data={data} title={title ?? data.title} />}
+      testId="seer-conversation-embed"
+    >
+      {error ? (
+        <Text variant="danger">{t('Unable to load conversation.')}</Text>
+      ) : !isLoading && nodes.length === 0 ? (
+        <Text variant="muted">{t('No messages in this conversation')}</Text>
+      ) : (
+        <ConversationAggregatesBar
+          conversationId={data.id}
+          nodes={nodes}
+          isLoading={isLoading}
+        />
+      )}
+    </QueryEmbedCard>
+  );
+}
diff --git a/static/app/components/seer/markdown/embeds/components/conversation/conversationLink.tsx b/static/app/components/seer/markdown/embeds/components/conversation/conversationLink.tsx
new file mode 100644
index 000000000000..3e5445bd2d1c
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversation/conversationLink.tsx
@@ -0,0 +1,73 @@
+import {ResourceLink} from 'sentry/components/seer/markdown/embeds/components/resourceLink';
+import type {EmbedOutput} from 'sentry/components/seer/markdown/embeds/utils';
+import {IconChat} from 'sentry/icons';
+import {t} from 'sentry/locale';
+import {normalizeUrl} from 'sentry/utils/url/normalizeUrl';
+import {useOrganization} from 'sentry/utils/useOrganization';
+import {
+  CONVERSATIONS_DETAIL_SUB_PATH,
+  EXPLORE_AGENTS_SUB_PATH,
+} from 'sentry/views/explore/conversations/settings';
+
+export type ConversationData = EmbedOutput<'conversation'>;
+
+/**
+ * The detail view scopes its span query to the URL's time range, so the window
+ * is padded either side of the conversation's own timestamps -- the same hour
+ * `getConversationDetailUrl` pads by for a row in the conversations table.
+ */
+const CONVERSATION_WINDOW_PADDING_MS = 60 * 60 * 1000;
+
+/**
+ * `getConversationDetailUrl` needs a full `Conversation` row, which an embed
+ * never has -- the tag carries an id and, at best, the conversation's first and
+ * last span timestamps. Build the same path from those instead.
+ */
+export function getConversationHref(
+  data: ConversationData,
+  organizationSlug: string,
+  referrer = 'seer-conversation-embed'
+): string {
+  const basePath = `/organizations/${organizationSlug}/explore/${EXPLORE_AGENTS_SUB_PATH}/${CONVERSATIONS_DETAIL_SUB_PATH}/${encodeURIComponent(data.id)}/`;
+
+  const params = new URLSearchParams();
+  if (data.start) {
+    params.set(
+      'start',
+      new Date(Date.parse(data.start) - CONVERSATION_WINDOW_PADDING_MS).toISOString()
+    );
+  }
+  if (data.end) {
+    params.set(
+      'end',
+      new Date(Date.parse(data.end) + CONVERSATION_WINDOW_PADDING_MS).toISOString()
+    );
+  }
+  for (const project of data.projects ?? []) {
+    params.append('project', String(project));
+  }
+  params.set('referrer', referrer);
+
+  return normalizeUrl(`${basePath}?${params.toString()}`);
+}
+
+interface ConversationLinkProps {
+  data: ConversationData;
+  /**
+   * Overrides the tag's title. The block passes the API-provided title once it
+   * has loaded, which is fresher than whatever the model wrote into the tag.
+   */
+  title?: string | null;
+}
+
+export function ConversationLink({data, title}: ConversationLinkProps) {
+  const organization = useOrganization();
+
+  return (
+    <ResourceLink
+      icon={IconChat}
+      href={getConversationHref(data, organization.slug)}
+      title={title ?? data.title ?? t('Conversation %s', data.id)}
+    />
+  );
+}
diff --git a/static/app/components/seer/markdown/embeds/components/conversationsQuery.spec.tsx b/static/app/components/seer/markdown/embeds/components/conversationsQuery.spec.tsx
new file mode 100644
index 000000000000..bb016ba8708f
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversationsQuery.spec.tsx
@@ -0,0 +1,180 @@
+import {act, screen, waitFor, within} from 'sentry-test/reactTestingLibrary';
+
+import {PageFiltersStore} from 'sentry/components/pageFilters/store';
+import {
+  getEmbedLinkHref,
+  renderEmbed,
+} from 'sentry/components/seer/markdown/embeds/components/resourceEmbedTestUtils';
+
+const LIST_URL = '/organizations/org-slug/agents/conversations/';
+
+/** Shaped as the list endpoint returns it: `firstInput` may be content parts. */
+function conversationFixture(overrides: Record<string, unknown> = {}) {
+  return {
+    conversationId: '8f0e1f6a-1f2b-4d3c-9e5a-0b1c2d3e4f5a',
+    title: 'Refund request escalated',
+    firstInput: [{type: 'text', text: 'I want a refund'}],
+    lastOutput: 'Escalating to a human',
+    duration: 120_000,
+    generationDuration: 4200,
+    llmCalls: 3,
+    toolCalls: 2,
+    toolErrors: 0,
+    toolNames: ['search'],
+    errors: 1,
+    inputTokens: 100,
+    outputTokens: 200,
+    totalTokens: 300,
+    totalCost: 0.42,
+    startTimestamp: 1_700_000_000_000,
+    endTimestamp: 1_700_000_120_000,
+    traceCount: 1,
+    traceIds: ['trace-1'],
+    projectId: 1,
+    user: {email: 'user@example.com', id: '1', ip_address: null, username: null},
+    ...overrides,
+  };
+}
+
+function searchParams(href: string) {
+  return new URLSearchParams(href.split('?')[1] ?? '');
+}
+
+describe('conversationsQuery embed', () => {
+  beforeAll(async () => {
+    // The block is behind `lazy()`; compile its module graph up front rather
+    // than inside the first `findBy*`.
+    await import('./conversationsQueryBlock');
+  }, 60_000);
+
+  beforeEach(() => {
+    MockApiClient.clearMockResponses();
+    act(() => {
+      PageFiltersStore.reset();
+      PageFiltersStore.init();
+    });
+  });
+
+  it('links to the agents list inline', () => {
+    const href = getEmbedLinkHref('conversationsQuery', 'Conversations using tools', {
+      query: 'gen_ai.tool.name:*',
+      statsPeriod: '24h',
+      projects: ['1', '2'],
+      environments: ['prod'],
+      agents: ['support-bot', 'triage-bot'],
+      title: 'Conversations using tools',
+    });
+
+    expect(href.split('?')[0]).toBe('/organizations/org-slug/explore/agents/');
+    const params = searchParams(href);
+    expect(params.get('query')).toBe('gen_ai.tool.name:*');
+    expect(params.get('statsPeriod')).toBe('24h');
+    expect(params.getAll('project')).toEqual(['1', '2']);
+    expect(params.getAll('environment')).toEqual(['prod']);
+    // The list reads `agent` as one comma-separated value.
+    expect(params.get('agent')).toBe('support-bot,triage-bot');
+    expect(params.get('referrer')).toBe('seer-conversations-query-embed');
+  });
+
+  it('falls back to a generic title', () => {
+    const href = getEmbedLinkHref('conversationsQuery', 'Conversation search', {
+      query: '',
+    });
+
+    expect(href).toContain('/organizations/org-slug/explore/agents/');
+  });
+
+  it('previews matching conversations', async () => {
+    const request = MockApiClient.addMockResponse({
+      url: LIST_URL,
+      body: [
+        conversationFixture({
+          conversationId: 'older',
+          title: 'Older conversation',
+          endTimestamp: 1_600_000_000_000,
+        }),
+        conversationFixture(),
+      ],
+    });
+
+    renderEmbed({
+      name: 'conversationsQuery',
+      data: {
+        query: 'gen_ai.request.model:gpt-4o',
+        statsPeriod: '24h',
+        agents: ['support-bot'],
+      },
+    });
+
+    expect(await screen.findByText('Refund request escalated')).toBeInTheDocument();
+
+    const row = screen.getByRole('row', {name: /Refund request escalated/});
+    // A UUID id is shown as a short prefix.
+    expect(within(row).getByText('8f0e1f6a')).toBeInTheDocument();
+    expect(within(row).getByText('user@example.com')).toBeInTheDocument();
+    expect(within(row).getByText('3')).toBeInTheDocument();
+    expect(within(row).getByText('1')).toBeInTheDocument();
+    expect(within(row).getByText('$0.42')).toBeInTheDocument();
+
+    // Newest-first, the way the list view orders its rows.
+    const rows = screen.getAllByRole('row');
+    expect(rows[1]).toHaveTextContent('Refund request escalated');
+    expect(rows[2]).toHaveTextContent('Older conversation');
+
+    await waitFor(() => {
+      expect(request).toHaveBeenCalledWith(
+        LIST_URL,
+        expect.objectContaining({
+          query: expect.objectContaining({
+            // The agent filter is a URL param on the list view, but part of
+            // the span query when the endpoint is called directly.
+            query: expect.stringContaining('support-bot'),
+            statsPeriod: '24h',
+            per_page: 5,
+          }),
+        })
+      );
+    });
+    expect(request.mock.calls[0][1].query.query).toContain('gen_ai.request.model:gpt-4o');
+  });
+
+  it('links each row to its conversation in a new tab', async () => {
+    MockApiClient.addMockResponse({url: LIST_URL, body: [conversationFixture()]});
+
+    renderEmbed({name: 'conversationsQuery', data: {query: ''}});
+
+    const link = await screen.findByRole('link', {name: 'Refund request escalated'});
+
+    // A new tab keeps the answer the embed is rendered into on screen.
+    expect(link).toHaveAttribute('target', '_blank');
+    const href = link.getAttribute('href') ?? '';
+    expect(href.split('?')[0]).toBe(
+      '/organizations/org-slug/explore/agents/conversations/8f0e1f6a-1f2b-4d3c-9e5a-0b1c2d3e4f5a/'
+    );
+    const params = searchParams(href);
+    // The row's own timestamps, padded the same hour the detail view expects.
+    expect(params.get('start')).toBe('2023-11-14T21:13:20.000Z');
+    expect(params.get('end')).toBe('2023-11-14T23:15:20.000Z');
+    expect(params.getAll('project')).toEqual(['1']);
+    expect(params.get('referrer')).toBe('seer-conversations-query-embed');
+  });
+
+  it('flattens a content-part first input when there is no title', async () => {
+    MockApiClient.addMockResponse({
+      url: LIST_URL,
+      body: [conversationFixture({title: null})],
+    });
+
+    renderEmbed({name: 'conversationsQuery', data: {query: ''}});
+
+    expect(await screen.findByText('I want a refund')).toBeInTheDocument();
+  });
+
+  it('shows an empty state when nothing matches', async () => {
+    MockApiClient.addMockResponse({url: LIST_URL, body: []});
+
+    renderEmbed({name: 'conversationsQuery', data: {query: 'gen_ai.tool.name:*'}});
+
+    expect(await screen.findByText('No matching conversations')).toBeInTheDocument();
+  });
+});
diff --git a/static/app/components/seer/markdown/embeds/components/conversationsQuery.tsx b/static/app/components/seer/markdown/embeds/components/conversationsQuery.tsx
new file mode 100644
index 000000000000..45583d2d403b
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversationsQuery.tsx
@@ -0,0 +1,18 @@
+import {lazy} from 'react';
+
+import {LazyLoad} from 'sentry/components/lazyLoad';
+import {defineSeerEmbed} from 'sentry/components/seer/markdown/embeds/utils';
+
+import {ConversationsQueryLink} from './conversationsQueryLink';
+
+const LazyConversationsQueryBlock = lazy(() => import('./conversationsQueryBlock'));
+
+export const ConversationsQuery = defineSeerEmbed({
+  name: 'conversationsQuery',
+  render(props, level) {
+    if (level === 'block') {
+      return <LazyLoad LazyComponent={LazyConversationsQueryBlock} data={props} />;
+    }
+    return <ConversationsQueryLink data={props} />;
+  },
+});
diff --git a/static/app/components/seer/markdown/embeds/components/conversationsQueryBlock.tsx b/static/app/components/seer/markdown/embeds/components/conversationsQueryBlock.tsx
new file mode 100644
index 000000000000..d76401e26775
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversationsQueryBlock.tsx
@@ -0,0 +1,222 @@
+import {useQuery} from '@tanstack/react-query';
+
+import {Flex, Stack} from '@sentry/scraps/layout';
+import {ExternalLink} from '@sentry/scraps/link';
+import {Text} from '@sentry/scraps/text';
+
+import {Count} from 'sentry/components/count';
+import {normalizeDateTimeParams} from 'sentry/components/pageFilters/parse';
+import {PerformanceDuration} from 'sentry/components/performanceDuration';
+import {QueryEmbedCard} from 'sentry/components/seer/markdown/embeds/components/queryEmbed/queryEmbedCard';
+import {QUERY_EMBED_ROW_LIMIT} from 'sentry/components/seer/markdown/embeds/components/queryEmbed/queryEmbedConstants';
+import {
+  QueryEmbedTable,
+  type QueryEmbedColumn,
+} from 'sentry/components/seer/markdown/embeds/components/queryEmbed/queryEmbedTable';
+import {toPageFilters} from 'sentry/components/seer/markdown/embeds/components/queryEmbedParams';
+import {t} from 'sentry/locale';
+import {apiOptions} from 'sentry/utils/api/apiOptions';
+import {markdownToPlainText} from 'sentry/utils/marked/marked';
+import {ellipsize} from 'sentry/utils/string/ellipsize';
+import {isUUID} from 'sentry/utils/string/isUUID';
+import {useOrganization} from 'sentry/utils/useOrganization';
+import type {
+  Conversation,
+  ConversationUser,
+} from 'sentry/views/explore/conversations/hooks/useConversations';
+import {LLMCosts} from 'sentry/views/insights/pages/agents/components/llmCosts';
+
+import {getConversationHref} from './conversation/conversationLink';
+import {
+  combineAgentQuery,
+  ConversationsQueryLink,
+  type ConversationsQueryData,
+} from './conversationsQueryLink';
+
+/**
+ * The list endpoint's raw rows. `useConversations` flattens the two content
+ * fields before handing them out, and this block calls the endpoint directly
+ * (that hook takes its filters from the router), so it flattens them itself.
+ */
+type ConversationApiRow = Omit<Conversation, 'firstInput' | 'lastOutput'> & {
+  firstInput?: Array<{text: string; type: string}> | string | null;
+  lastOutput?: Array<{text: string; type: string}> | string | null;
+};
+
+/** A preview line is one table cell wide, so it is cut well short of the table's limit. */
+const PREVIEW_MAX_CHARS = 200;
+
+function flattenContent(content: ConversationApiRow['firstInput']): string | null {
+  if (typeof content === 'string') {
+    return content;
+  }
+  return content?.find(part => part.type === 'text')?.text ?? null;
+}
+
+/**
+ * A title is model-written markdown and a first input is whatever the user
+ * typed, so both are flattened to a single line before they go in a cell.
+ */
+function getConversationLabel(row: ConversationApiRow): string | null {
+  const raw = row.title ?? flattenContent(row.firstInput);
+  if (!raw) {
+    return null;
+  }
+  const plainText = ellipsize(
+    markdownToPlainText(raw).replace(/\s+/g, ' ').trim(),
+    PREVIEW_MAX_CHARS
+  );
+  return plainText.length > 0 ? plainText : null;
+}
+
+/** `useConversations` renders the SDK's literal "none" as no user at all. */
+function getUserLabel(user: ConversationUser | null): string | null {
+  const fields = [user?.email, user?.username, user?.ip_address, user?.id];
+  return fields.find(value => value && value.toLowerCase() !== 'none') ?? null;
+}
+
+/** A row carries epoch milliseconds; the detail URL wants ISO timestamps. */
+function toIsoTimestamp(timestamp: number | null | undefined): string | undefined {
+  return timestamp ? new Date(timestamp).toISOString() : undefined;
+}
+
+function getConversationIdLabel(conversationId: string): string {
+  // UUIDs are long and opaque, so show a short prefix; other id formats
+  // (e.g. `resp_...`, `slack:1234`) are already short enough.
+  return isUUID(conversationId) ? conversationId.slice(0, 8) : conversationId;
+}
+
+/**
+ * Opens in a new tab so following a row cannot replace the page the embed is
+ * rendered into -- the answer above it would be lost. `getConversationHref`
+ * builds the same detail URL the `conversation` embed links to.
+ */
+function ConversationCell({row}: {row: ConversationApiRow}) {
+  const organization = useOrganization();
+  const label = getConversationLabel(row);
+  const user = getUserLabel(row.user);
+
+  return (
+    <Stack gap="xs" minWidth={0}>
+      <Text ellipsis>
+        <ExternalLink
+          href={getConversationHref(
+            {
+              id: row.conversationId,
+              start: toIsoTimestamp(row.startTimestamp),
+              end: toIsoTimestamp(row.endTimestamp),
+              projects: row.projectId === null ? undefined : [String(row.projectId)],
+            },
+            organization.slug,
+            'seer-conversations-query-embed'
+          )}
+        >
+          {label ?? t('Untitled conversation')}
+        </ExternalLink>
+      </Text>
+      <Flex gap="sm" minWidth={0}>
+        <Text size="sm" variant="muted" ellipsis>
+          {getConversationIdLabel(row.conversationId)}
+        </Text>
+        {user ? (
+          <Text size="sm" variant="muted" ellipsis>
+            {user}
+          </Text>
+        ) : null}
+      </Flex>
+    </Stack>
+  );
+}
+
+const COLUMNS: Array<QueryEmbedColumn<ConversationApiRow>> = [
+  {
+    key: 'conversation',
+    label: t('Conversation'),
+    render: row => <ConversationCell row={row} />,
+  },
+  {
+    key: 'duration',
+    label: t('Duration'),
+    // The generation duration, not the wall-clock span of the conversation:
+    // a conversation can sit idle for hours between two messages.
+    render: row => (
+      <Text ellipsis tabular>
+        <PerformanceDuration milliseconds={row.generationDuration} abbreviation />
+      </Text>
+    ),
+  },
+  {
+    key: 'messages',
+    label: t('Messages'),
+    render: row => (
+      <Text ellipsis tabular>
+        <Count value={row.llmCalls} />
+      </Text>
+    ),
+  },
+  {
+    key: 'errors',
+    label: t('Errors'),
+    render: row => (
+      <Text ellipsis tabular variant={row.errors > 0 ? 'danger' : undefined}>
+        <Count value={row.errors} />
+      </Text>
+    ),
+  },
+  {
+    key: 'cost',
+    label: t('Cost'),
+    render: row => (
+      <Text ellipsis tabular>
+        <LLMCosts cost={row.totalCost} />
+      </Text>
+    ),
+  },
+];
+
+export default function ConversationsQueryBlock({data}: {data: ConversationsQueryData}) {
+  const organization = useOrganization();
+  const selection = toPageFilters(data);
+
+  const conversationsQuery = useQuery({
+    ...apiOptions.as<ConversationApiRow[]>()(
+      '/organizations/$organizationIdOrSlug/agents/conversations/',
+      {
+        path: {organizationIdOrSlug: organization.slug},
+        query: {
+          query: combineAgentQuery(data.query, data.agents),
+          project: selection.projects,
+          environment: selection.environments,
+          per_page: QUERY_EMBED_ROW_LIMIT,
+          ...normalizeDateTimeParams(selection.datetime),
+        },
+        staleTime: 30_000,
+      }
+    ),
+    retry: false,
+  });
+
+  // The endpoint orders by relevance rather than recency; the list view sorts
+  // newest-first before rendering, so the preview shows the same five rows.
+  const rows = (conversationsQuery.data ?? []).toSorted(
+    (a, b) => b.endTimestamp - a.endTimestamp
+  );
+
+  return (
+    <QueryEmbedCard
+      link={<ConversationsQueryLink data={data} />}
+      query={data.query}
+      testId="seer-conversations-query-embed"
+    >
+      <QueryEmbedTable
+        columns={COLUMNS}
+        emptyMessage={t('No matching conversations')}
+        errorMessage={t('Unable to load conversations')}
+        isError={conversationsQuery.isError}
+        isPending={conversationsQuery.isPending}
+        rowKey={row => row.conversationId}
+        rows={rows}
+      />
+    </QueryEmbedCard>
+  );
+}
diff --git a/static/app/components/seer/markdown/embeds/components/conversationsQueryLink.tsx b/static/app/components/seer/markdown/embeds/components/conversationsQueryLink.tsx
new file mode 100644
index 000000000000..210a96be6a65
--- /dev/null
+++ b/static/app/components/seer/markdown/embeds/components/conversationsQueryLink.tsx
@@ -0,0 +1,67 @@
+import queryString from 'query-string';
+
+import {ResourceLink} from 'sentry/components/seer/markdown/embeds/components/resourceLink';
+import type {EmbedOutput} from 'sentry/components/seer/markdown/embeds/utils';
+import {IconChat} from 'sentry/icons';
+import {t} from 'sentry/locale';
+import type {Organization} from 'sentry/types/organization';
+import {normalizeUrl} from 'sentry/utils/url/normalizeUrl';
+import {useOrganization} from 'sentry/utils/useOrganization';
+import {EXPLORE_AGENTS_SUB_PATH} from 'sentry/views/explore/conversations/settings';
+import {getAgentNamesFilter} from 'sentry/views/insights/pages/agents/utils/query';
+
+export type ConversationsQueryData = EmbedOutput<'conversationsQuery'>;
+
+/**
+ * The agents list keeps the agent filter in its own `agent` param and folds it
+ * into the span query itself only when it calls the API -- see
+ * `useCombinedQuery`. Mirror both halves so the link and the block preview
+ * below it are filtered the same way.
+ */
+export function combineAgentQuery(query: string, agents?: string[]): string {
+  const agentQuery = getAgentNamesFilter(agents ?? []);
+  if (!agentQuery) {
+    return query;
+  }
+  if (!query) {
+    return agentQuery;
+  }
+  return `(${agentQuery}) and (${query})`;
+}
+
+function getConversationsQueryHref(
+  data: ConversationsQueryData,
+  organization: Organization
+): string {
+  const {query, agents, projects, environments, statsPeriod, start, end} = data;
+
+  return queryString.stringifyUrl({
+    url: normalizeUrl(
+      `/organizations/${organization.slug}/explore/${EXPLORE_AGENTS_SUB_PATH}/`
+    ),
+    query: {
+      query,
+      project: projects,
+      environment: environments,
+      statsPeriod,
+      start,
+      end,
+      // The list reads `agent` as a single comma-separated value, not repeated
+      // params.
+      agent: agents?.length ? agents.join(',') : undefined,
+      referrer: 'seer-conversations-query-embed',
+    },
+  });
+}
+
+export function ConversationsQueryLink({data}: {data: ConversationsQueryData}) {
+  const organization = useOrganization();
+
+  return (
+    <ResourceLink
+      icon={IconChat}
+      href={getConversationsQueryHref(data, organization)}
+      title={data.title ?? t('Conversation search')}
+    />
+  );
+}
diff --git a/static/app/components/seer/markdown/embeds/index.ts b/static/app/components/seer/markdown/embeds/index.ts
index ca31b55e1875..f54556af2dd1 100644
--- a/static/app/components/seer/markdown/embeds/index.ts
+++ b/static/app/components/seer/markdown/embeds/index.ts
@@ -2,6 +2,8 @@ import {AgentWriteApprovalEmbed} from './components/agentWriteApproval';
 import {Alert} from './components/alert/alert';
 import {Autofix, AutofixRef} from './components/autofix';
 import {Chart} from './components/chart';
+import {Conversation} from './components/conversation/conversation';
+import {ConversationsQuery} from './components/conversationsQuery';
 import {Dashboard} from './components/dashboard';
 import {Docs} from './components/docs';
 import {Dsn} from './components/dsn';
@@ -29,6 +31,8 @@ const embeds = [
   Autofix,
   AutofixRef,
   Chart,
+  Conversation,
+  ConversationsQuery,
   Dashboard,
   Docs,
   Dsn,
diff --git a/static/app/components/seer/markdown/embeds/schemas.ts b/static/app/components/seer/markdown/embeds/schemas.ts
index 8f9f369d7ec3..dd567a81a3d1 100644
--- a/static/app/components/seer/markdown/embeds/schemas.ts
+++ b/static/app/components/seer/markdown/embeds/schemas.ts
@@ -680,6 +680,70 @@ export const SEER_EMBED_SCHEMAS = {
       },
     ],
   },
+  conversation: {
+    description:
+      'The ONLY way to reference a single AI agent conversation (Explore > Agents). ' +
+      'Use the `conversationId` exactly as the agents conversations API returns it. ' +
+      'Include the API-provided `title` when available, and `start`/`end` (the ' +
+      "conversation's own first and last span timestamps) so the embed can scope " +
+      'its query instead of scanning the default window. ' +
+      'Inline: renders a compact link. ' +
+      'Block: renders the conversation transcript with its LLM call, token, cost, ' +
+      'and tool totals. Do not duplicate the messages or those totals as text. ' +
+      'Never use a markdown link for conversation references.',
+    featureFlag: 'organizations:gen-ai-conversations',
+    level: ['inline', 'block'],
+    schema: z.object({
+      id: z.string().min(1),
+      title: z.string().min(1).optional(),
+      projects: z.array(idString).optional(),
+      start: isoTimestampSchema.optional(),
+      end: isoTimestampSchema.optional(),
+    }),
+    examples: [
+      {
+        label: 'Conversation',
+        data: {
+          id: '4821',
+          title: 'Refund request escalated to a human',
+          start: '2026-08-25T16:37:12Z',
+          end: '2026-08-25T16:39:02Z',
+        },
+      },
+    ],
+  },
+  conversationsQuery: {
+    description:
+      'Preview the AI agent conversations list (Explore > Agents) filtered by a ' +
+      'search query. Use this when pointing the user at a SET of conversations — ' +
+      'if you have a specific conversation ID, use the `conversation` embed instead. ' +
+      '`query` uses span search syntax over gen_ai spans, e.g. ' +
+      '"gen_ai.request.model:gpt-4o". Negation is not supported. ' +
+      'Use `agents` to filter to specific agent names. ' +
+      'Inline renders a link; block renders the first five matching conversations ' +
+      'with their duration, message count, errors and cost.',
+    featureFlag: 'organizations:gen-ai-conversations',
+    level: ['inline', 'block'],
+    schema: z.object({
+      ...pageFilterFields,
+      query: z.string().default(''),
+      agents: z
+        .array(z.string())
+        .optional()
+        .describe('Filter to these agent names, as reported by gen_ai.agent.name.'),
+      title: z.string().min(1).optional(),
+    }),
+    examples: [
+      {
+        label: 'Conversations with tool errors',
+        data: {
+          query: 'gen_ai.tool.name:*',
+          statsPeriod: '24h',
+          title: 'Conversations using tools',
+        },
+      },
+    ],
+  },
   replaysQuery: {
     description:
       'Preview the Session Replay list filtered by a search query. ' +
diff --git a/static/app/components/seer/markdown/seerMarkdown.mdx b/static/app/components/seer/markdown/seerMarkdown.mdx
index 76226ee2ccf8..7280e8c99e90 100644
--- a/static/app/components/seer/markdown/seerMarkdown.mdx
+++ b/static/app/components/seer/markdown/seerMarkdown.mdx
@@ -10,6 +10,7 @@ import * as Storybook from 'sentry/stories';
 
 import {BasicDemo, LinkifyDemo, StreamingEmbedExamples} from './__stories__/components';
 import {AlertEmbedStory} from './__stories__/alertEmbedStory';
+import {ConversationEmbedStory} from './__stories__/conversationEmbedStory';
 import {DashboardEmbedStory} from './__stories__/dashboardEmbedStory';
 import {EmbedStory} from './__stories__/embedStory';
 import {MonitorEmbedStory} from './__stories__/monitorEmbedStory';
@@ -137,6 +138,14 @@ Tag syntax: `{% name %}{"key":"value"}{% /name %}`. The JSON body is validated a
 
 <EmbedStory name="metricsQuery" />
 
+### conversation
+
+<ConversationEmbedStory />
+
+### conversationsQuery
+
+<EmbedStory name="conversationsQuery" />
+
 ## Embed Architecture
 
 The embed system has four parts:
diff --git a/static/app/views/explore/conversations/hooks/useConversation.tsx b/static/app/views/explore/conversations/hooks/useConversation.tsx
index 9459f263d2df..a565adf0d13e 100644
--- a/static/app/views/explore/conversations/hooks/useConversation.tsx
+++ b/static/app/views/explore/conversations/hooks/useConversation.tsx
@@ -19,6 +19,13 @@ import type {TraceTree} from 'sentry/views/performance/newTraceDetails/traceMode
 export interface UseConversationsOptions {
   conversationId: string;
   endTimestamp?: number;
+  /**
+   * Projects to scope the span query to, overriding the page filters. A caller
+   * that is not the conversations route -- an embed rendered into some other
+   * page -- knows the conversation's own project and must not inherit whatever
+   * the host page happens to have selected.
+   */
+  projects?: number[];
   startTimestamp?: number;
 }
 
@@ -314,8 +321,8 @@ export function useConversation(
       ? normalizeDateTimeParams(selection.datetime)
       : {};
 
-  const project =
-    selection.projects.length > 0 ? selection.projects : [ALL_ACCESS_PROJECTS];
+  const selectedProjects = conversation.projects ?? selection.projects;
+  const project = selectedProjects.length > 0 ? selectedProjects : [ALL_ACCESS_PROJECTS];
 
   const queryParams = {
     project,