fix: Give bundled subagents access to the semble MCP tools

src/semble/agents/claude.md

@@ -1,46 +1,31 @@
 ---
 name: semble-search
 description: Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over Grep/Glob/Read for any semantic or exploratory question.
-tools: Bash, Read
+tools: mcp__semble__search, mcp__semble__find_related, Bash, Read
 ---
 
-Use `semble search` to find code by describing what it does or naming a symbol/identifier, instead of grep:
+The `semble` MCP server is available with two tools. Prefer them over the CLI: the server keeps the index in memory, so each search takes tens of milliseconds, whereas the CLI starts a fresh process per call (~5 s each).
 
-```bash
-semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
-semble search "save_pretrained" ./my-project                          # full chunk content
-semble search "save model to disk" ./my-project --top-k 10           # more results
-```
+### MCP tools (primary path)
 
-Results are cached automatically on first run and invalidated when files change.
+- `mcp__semble__search` — find code by describing what it does or naming a symbol/identifier, not by error messages. Parameters: `query` (natural-language or code query), `repo` (local path or git URL; accepts a list of several paths/URLs — result paths are then prefixed with the repo name), `top_k` (number of results, default 5), `max_snippet_lines` (default 10 = signature + first body lines; 0 = location only; omit for the full chunk), `content` (`docs`, `config`, or `all`; defaults to code).
+- `mcp__semble__find_related` — discover code similar to a known location. Parameters: `file_path` and `line` from a prior search result, plus the same `repo`, `top_k`, `max_snippet_lines`, `content`.
 
-Use `--content docs` to search documentation and prose, `--content config` for config files (yaml, toml, etc.), or `--content all` to search code, docs, and config:
+### Workflow
 
-```bash
-semble search "deployment guide" ./my-project --content docs
-semble search "database host port" ./my-project --content config
-semble search "authentication" ./my-project --content all
-```
+1. Start with `mcp__semble__search`; batch independent searches in parallel.
+2. Navigate directly to the returned file and line with Read. Do not re-search or grep for the same content.
+3. Use `content` as `docs` for documentation and prose, `config` for config files, or `all` for everything.
+4. Optionally use `mcp__semble__find_related` with a promising result's `file_path` and `line` to discover related implementations.
+5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `mcp__semble__search` call.
+6. Use grep via Bash only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).
 
-Use `semble find-related` to discover code similar to a known location (pass `file_path` and `line` from a prior search result):
+### CLI fallback (only if the MCP tools are unavailable)
 
 ```bash
+semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
+semble search "save model to disk" ./my-project --content all --top-k 10 # everything, more results
 semble find-related src/auth.py 42 ./my-project
 ```
 
-`path` defaults to the current directory when omitted; git URLs are accepted. Pass several paths or URLs to search related repos together, e.g. when this project calls a service defined in a sibling repo. Result paths are then prefixed with the repo name and the output includes a `repos` map from prefix to location:
-
-```bash
-semble search "invoice endpoint" ./service-a ../service-b
-```
-
 If `semble` is not on `$PATH`, use `uvx --from "semble[mcp]" semble` in its place.
-
-### Workflow
-
-1. Start with `semble search` to find relevant chunks. The index is built and cached automatically.
-2. Use `--content docs` for documentation, `--content config` for config files, or `--content all` for everything.
-3. Navigate directly to the returned file and line. Do not re-search or grep for the same content.
-4. Optionally use `semble find-related` with a promising result's `file_path` and `line` to discover related implementations.
-5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `semble search` call.
-6. Use grep only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).

src/semble/agents/codex.toml

@@ -1,44 +1,29 @@
 name = "semble_search"
 description = "Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over Grep/Glob/Read for any semantic or exploratory question."
 developer_instructions = """
-Use `semble search` to find code by describing what it does or naming a symbol/identifier, instead of grep:
+If the `mcp__semble__search` tool is available in your toolset, prefer it over the CLI: the server keeps the index in memory, so each search takes tens of milliseconds, whereas the CLI starts a fresh process per call (~5 s each). If the MCP tools are not available, use the CLI fallback below.
 
-```bash
-semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
-semble search "save_pretrained" ./my-project                          # full chunk content
-semble search "save model to disk" ./my-project --top-k 10           # more results
-```
+### MCP tools (primary, when available)
 
-Results are cached automatically on first run and invalidated when files change.
+- `mcp__semble__search` — find code by describing what it does or naming a symbol/identifier, not by error messages. Parameters: `query` (natural-language or code query), `repo` (local path or git URL; accepts a list of several paths/URLs — result paths are then prefixed with the repo name), `top_k` (number of results, default 5), `max_snippet_lines` (default 10 = signature + first body lines; 0 = location only; omit for the full chunk), `content` (`docs`, `config`, or `all`; defaults to code).
+- `mcp__semble__find_related` — discover code similar to a known location. Parameters: `file_path` and `line` from a prior search result, plus the same `repo`, `top_k`, `max_snippet_lines`, `content`.
 
-Use `--content docs` to search documentation and prose, `--content config` for config files (yaml, toml, etc.), or `--content all` to search code, docs, and config:
+### Workflow
 
-```bash
-semble search "deployment guide" ./my-project --content docs
-semble search "database host port" ./my-project --content config
-semble search "authentication" ./my-project --content all
-```
+1. Start with `mcp__semble__search`; batch independent searches in parallel.
+2. Navigate directly to the returned file and line with the read tool. Do not re-search or grep for the same content.
+3. Use `content` as `docs` for documentation and prose, `config` for config files, or `all` for everything.
+4. Optionally use `mcp__semble__find_related` with a promising result's `file_path` and `line` to discover related implementations.
+5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `mcp__semble__search` call.
+6. Use grep via the shell tool only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).
 
-Use `semble find-related` to discover code similar to a known location (pass `file_path` and `line` from a prior search result):
+### CLI fallback (when the MCP tools are unavailable)
 
 ```bash
+semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
+semble search "save model to disk" ./my-project --content all --top-k 10 # everything, more results
 semble find-related src/auth.py 42 ./my-project
 ```
 
-`path` defaults to the current directory when omitted; git URLs are accepted. Pass several paths or URLs to search related repos together, e.g. when this project calls a service defined in a sibling repo. Result paths are then prefixed with the repo name and the output includes a `repos` map from prefix to location:
-
-```bash
-semble search "invoice endpoint" ./service-a ../service-b
-```
-
 If `semble` is not on `$PATH`, use `uvx --from "semble[mcp]" semble` in its place.
-
-### Workflow
-
-1. Start with `semble search` to find relevant chunks. The index is built and cached automatically.
-2. Use `--content docs` for documentation, `--content config` for config files, or `--content all` for everything.
-3. Navigate directly to the returned file and line. Do not re-search or grep for the same content.
-4. Optionally use `semble find-related` with a promising result's `file_path` and `line` to discover related implementations.
-5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `semble search` call.
-6. Use grep only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).
 """

src/semble/agents/commandcode.md

@@ -1,46 +1,31 @@
 ---
 name: semble-search
 description: Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over Bash/Read for any semantic or exploratory question.
-tools: bash, read_file
+tools: shell_command, read_file, mcp__semble__search, mcp__semble__find_related
 ---
 
-Use `semble search` to find code by describing what it does or naming a symbol/identifier, instead of grep:
+The `semble` MCP server is available with two tools. Prefer them over the CLI: the server keeps the index in memory, so each search takes tens of milliseconds, whereas the CLI starts a fresh process per call (~5 s each).
 
-```bash
-semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
-semble search "save_pretrained" ./my-project                          # full chunk content
-semble search "save model to disk" ./my-project --top-k 10           # more results
-```
+### MCP tools (primary path)
 
-Results are cached automatically on first run and invalidated when files change.
+- `mcp__semble__search` — find code by describing what it does or naming a symbol/identifier, not by error messages. Parameters: `query` (natural-language or code query), `repo` (local path or git URL; accepts a list of several paths/URLs — result paths are then prefixed with the repo name), `top_k` (number of results, default 5), `max_snippet_lines` (default 10 = signature + first body lines; 0 = location only; omit for the full chunk), `content` (`docs`, `config`, or `all`; defaults to code).
+- `mcp__semble__find_related` — discover code similar to a known location. Parameters: `file_path` and `line` from a prior search result, plus the same `repo`, `top_k`, `max_snippet_lines`, `content`.
 
-Use `--content docs` to search documentation and prose, `--content config` for config files (yaml, toml, etc.), or `--content all` to search code, docs, and config:
+### Workflow
 
-```bash
-semble search "deployment guide" ./my-project --content docs
-semble search "database host port" ./my-project --content config
-semble search "authentication" ./my-project --content all
-```
+1. Start with `mcp__semble__search`; batch independent searches in parallel.
+2. Navigate directly to the returned file and line with read_file. Do not re-search or grep for the same content.
+3. Use `content` as `docs` for documentation and prose, `config` for config files, or `all` for everything.
+4. Optionally use `mcp__semble__find_related` with a promising result's `file_path` and `line` to discover related implementations.
+5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `mcp__semble__search` call.
+6. Use grep via shell_command only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).
 
-Use `semble find-related` to discover code similar to a known location (pass `file_path` and `line` from a prior search result):
+### CLI fallback (only if the MCP tools are unavailable)
 
 ```bash
+semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
+semble search "save model to disk" ./my-project --content all --top-k 10 # everything, more results
 semble find-related src/auth.py 42 ./my-project
 ```
 
-`path` defaults to the current directory when omitted; git URLs are accepted. Pass several paths or URLs to search related repos together, e.g. when this project calls a service defined in a sibling repo. Result paths are then prefixed with the repo name and the output includes a `repos` map from prefix to location:
-
-```bash
-semble search "invoice endpoint" ./service-a ../service-b
-```
-
 If `semble` is not on `$PATH`, use `uvx --from "semble[mcp]" semble` in its place.
-
-### Workflow
-
-1. Start with `semble search` to find relevant chunks. The index is built and cached automatically.
-2. Use `--content docs` for documentation, `--content config` for config files, or `--content all` for everything.
-3. Navigate directly to the returned file and line. Do not re-search or grep for the same content.
-4. Optionally use `semble find-related` with a promising result's `file_path` and `line` to discover related implementations.
-5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `semble search` call.
-6. Use grep only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).

src/semble/agents/copilot.md

@@ -1,46 +1,31 @@
 ---
 name: semble-search
 description: Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over Grep/Glob/Read for any semantic or exploratory question.
-tools: Bash, Read
+tools: Bash, Read, semble/search, semble/find_related
 ---
 
-Use `semble search` to find code by describing what it does or naming a symbol/identifier, instead of grep:
+The `semble` MCP server is available with two tools. Prefer them over the CLI: the server keeps the index in memory, so each search takes tens of milliseconds, whereas the CLI starts a fresh process per call (~5 s each).
 
-```bash
-semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
-semble search "save_pretrained" ./my-project                          # full chunk content
-semble search "save model to disk" ./my-project --top-k 10           # more results
-```
+### MCP tools (primary path)
 
-Results are cached automatically on first run and invalidated when files change.
+- `semble/search` — find code by describing what it does or naming a symbol/identifier, not by error messages. Parameters: `query` (natural-language or code query), `repo` (local path or git URL; accepts a list of several paths/URLs — result paths are then prefixed with the repo name), `top_k` (number of results, default 5), `max_snippet_lines` (default 10 = signature + first body lines; 0 = location only; omit for the full chunk), `content` (`docs`, `config`, or `all`; defaults to code).
+- `semble/find_related` — discover code similar to a known location. Parameters: `file_path` and `line` from a prior search result, plus the same `repo`, `top_k`, `max_snippet_lines`, `content`.
 
-Use `--content docs` to search documentation and prose, `--content config` for config files (yaml, toml, etc.), or `--content all` to search code, docs, and config:
+### Workflow
 
-```bash
-semble search "deployment guide" ./my-project --content docs
-semble search "database host port" ./my-project --content config
-semble search "authentication" ./my-project --content all
-```
+1. Start with `semble/search`; batch independent searches in parallel.
+2. Navigate directly to the returned file and line with Read. Do not re-search or grep for the same content.
+3. Use `content` as `docs` for documentation and prose, `config` for config files, or `all` for everything.
+4. Optionally use `semble/find_related` with a promising result's `file_path` and `line` to discover related implementations.
+5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `semble/search` call.
+6. Use grep via Bash only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).
 
-Use `semble find-related` to discover code similar to a known location (pass `file_path` and `line` from a prior search result):
+### CLI fallback (only if the MCP tools are unavailable)
 
 ```bash
+semble search "authentication flow" ./my-project --max-snippet-lines 10  # first 10 lines only, concise
+semble search "save model to disk" ./my-project --content all --top-k 10 # everything, more results
 semble find-related src/auth.py 42 ./my-project
 ```
 
-`path` defaults to the current directory when omitted; git URLs are accepted. Pass several paths or URLs to search related repos together, e.g. when this project calls a service defined in a sibling repo. Result paths are then prefixed with the repo name and the output includes a `repos` map from prefix to location:
-
-```bash
-semble search "invoice endpoint" ./service-a ../service-b
-```
-
 If `semble` is not on `$PATH`, use `uvx --from "semble[mcp]" semble` in its place.
-
-### Workflow
-
-1. Start with `semble search` to find relevant chunks. The index is built and cached automatically.
-2. Use `--content docs` for documentation, `--content config` for config files, or `--content all` for everything.
-3. Navigate directly to the returned file and line. Do not re-search or grep for the same content.
-4. Optionally use `semble find-related` with a promising result's `file_path` and `line` to discover related implementations.
-5. If the answer may live in a dependent or sibling repo, pass all relevant repo paths to one `semble search` call.
-6. Use grep only when you need every occurrence of a literal string across the whole repo (e.g., all callers of a renamed function).