Codesearch
flupkede103 顆星
Multi-repo semantic code search MCP server in Rust — hybrid vector + BM25 retrieval, tree-sitter AST chunking, fully offline. For OpenCode, Claude Code, Cursor, and any MCP client.
- 傳輸方式
- stdio
- 狀態
- 本機
接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令？ 兩種安裝方式的差異
Multi-repo semantic code search for AI agents — a Rust MCP server with vector + BM25 hybrid retrieval, symbol navigation, and cross-repository orchestration. Fully local, fully offline, no GPU, no Docker.
codesearch gives AI agents (OpenCode, Claude Code, Cursor, and any MCP client) deep codebase understanding through 5 unified MCP tools. Index once, search semantically across multiple repositories simultaneously.
get_chunkcodesearch index && codesearch mcp — doneThe MCP code-search ecosystem grew rapidly in late 2025 / early 2026 and many projects share the same baseline stack (Rust + tree-sitter + BM25 + embeddings + MCP). codesearch's deliberate focus is:
codesearch is intentionally narrower than full code-graph or knowledge-graph tools — it picks "lightweight, multi-repo, MCP-native, fully offline" and stays on that lane.
graph TB
Agent[AI Agent / MCP Client] -->|MCP stdio or HTTP| Router{MCP Router}
Router --> Search[search tool]
Router --> Find[find tool]
Router --> Explore[explore tool]
Router --> GetChunk[get_chunk tool]
Router --> FindImpact[find_impact tool]
Router --> Status[status tool]
Search -->|mode=semantic| Semantic[Vector ANN + BM25 + RRF Fusion]
Search -->|mode=literal| Literal[Tantivy FTS / Regex]
Find -->|definition/usages| SymbolIndex[Symbol Index]
Find -->|imports/dependents| DepGraph[Dependency Graph]
Explore -->|outline| TreeSitter[Tree-sitter AST]
Explore -->|similar| Semantic
Semantic --> Arroy[arroy ANN vectors]
Semantic --> Tantivy[Tantivy BM25]
Arroy --> LMDB[(LMDB)]
Tantivy --> TantivyIdx[(Tantivy Index)]
GetChunk --> LMDB
FindImpact -->|C# symbols| CSharpHelper[scip-csharp helper]
CSharpHelper -->|SCIP index| ScipLMDB[(LMDB scip_symbols)]
subgraph "Serve Mode (multi-repo + federation)"
ServeRouter[HTTP Router] -->|project/group routing| Repo1[Repo A]
ServeRouter --> Repo2[Repo B]
ServeRouter --> RepoN[Repo N]
ServeRouter -->|"@peer fan-out · TLS"| CloudPeer["Cloud serve peer<br/>results merged via RRF"]
end
Router -->|client mode| ServeRouter
Download pre-built binaries from Releases:
Or build from source:
git clone https://github.com/flupkede/codesearch.git
cd codesearch
cargo build --release
# Register and index the current repo (adds to ~/.codesearch/repos.json)
codesearch index add
# Register and index a repo from outside the repo folder
codesearch index add /path/to/my-project
# Incremental update (only changed files)
codesearch index /path/to/my-project
# Full rebuild
codesearch index /path/to/my-project --force
codesearch index add is intended to be run from inside the repo you want to register — pass the path explicitly if launched from elsewhere. First-time indexing takes 2–5 minutes; subsequent runs are incremental (10–30s) and branch switches re-index automatically. Use codesearch index list/rm/prune to manage registrations (see Serve Mode).
The default quantized MiniLM model favors startup speed and a small download. For multilingual text and notes, EmbeddingGemma 300M is available as a quantized ONNX model with retrieval-specific query and document prompts:
codesearch --model embeddinggemma-q4 index /path/to/notes --force
Changing models requires a full reindex because embedding dimensions and vector
spaces are model-specific. Keep the same model selected for later indexing runs.
Search rejects a --model value that differs from the indexed model and points
to the required --force rebuild instead of mixing incompatible vector spaces.
In serve mode the model is resolved per repository, from each index's own
metadata, not from a hub-wide setting: a hub may hold indexes built with
different models, and every query is embedded with the model of the repo it
targets (mixed-model groups are fine). codesearch serve --model <name> sets a
default for newly created indexes: a repo added without an explicit model
(e.g. codesearch index add with no --model, delegated to serve) is indexed
with it, and it is reported in GET /status as default_model. It never
overrides an index that already records its own model — to change an existing
repo's model, re-index that repo
(codesearch --model <name> index <path> --force) and restart serve. A repo
whose metadata.json records no model (a legacy index built before the
recording contract) is queried with the built-in 384-dim default rather than the
serve default, and the search response carries a warning naming the assumed model
and the re-index command.
codesearch connects to AI agents via MCP. Two modes:
The agent spawns codesearch mcp as a subprocess. It auto-detects the nearest index and starts a file watcher.
OpenCode — ~/.config/opencode/config.json:
{
"mcp": {
"codesearch": {
"type": "local",
"command": ["codesearch", "mcp"],
"enabled": true
}
}
}
Claude Code / Claude Desktop — ~/.config/claude-code/config.json or claude_desktop_config.json (identical schema):
{
"mcpServers": {
"codesearch": {
"command": "codesearch",
"args": ["mcp"]
}
}
}
Start the server first, then connect your agent. The server manages all registered repos with a TUI dashboard, lazy filesystem watchers, and idle eviction.
# Start the server (default port 39725)
codesearch serve
OpenCode — connect via HTTP:
{
"mcp": {
"codesearch": {
"type": "remote",
"url": "http://127.0.0.1:39725/mcp",
"enabled": true
}
}
}
Claude Code / Claude Desktop — force serve connection via --mode client:
{
"mcpServers": {
"codesearch": {
"command": "codesearch",
"args": ["mcp", "--mode", "client"]
}
}
}
Note: In multi-repo mode, agents must specify
projectorgroupin tool calls.statusalways works without scope.get_chunkauto-routes when the chunk_id is unique across repos; if ambiguous, it returns candidates and requiresproject.
codesearch publishes instructions to every MCP client on connect (via the initialize handshake) — when to reach for codesearch vs grep/glob, which tool to pick, and the serve-mode caveats. Most clients (OpenCode, Cursor) surface these automatically.
If your agent skips codesearch and falls back to grep/glob too often, paste this quickstart into its rules (AGENTS.md for Claude Code/OpenCode, .cursorrules for Cursor):
Prefer codesearch for semantic, cross-file, or symbol-oriented lookup ("where is X implemented", "find usages of Y", "how does Z flow"). Use plain grep/glob for a single known file, trivial one-line edits, or exact literal searches. In remote-serve mode, returned paths are from the server's filesystem — read content via
get_chunkrather than opening paths locally, and unindexed dirs (.venv,node_modules,build/) simply return nothing.
OpenCode: put this in the user-level ~/.config/opencode/AGENTS.md (applies across all projects), or install the OpenCode plugin, which injects this guidance with the resolved project scope automatically. Claude Code reads a project-level AGENTS.md, so add it per-project (or symlink a shared one).
Claude Code specifically tends to ignore this advice more than other clients — its MCP tool schemas are deferred (an extra ToolSearch call is needed before codesearch tools are even callable), while Grep/Glob are always fully loaded and zero-friction, and spawned subagents don't inherit AGENTS.md or the MCP initialize instructions at all.
To make the preference structural instead of advisory, this repo ships five Claude Code hooks (four PreToolUse guards plus one PostToolUse companion):
grep-guard — on Grep. Blocks a grep against an in-repo path when codesearch covers that repo (the target repo — resolved from the grep target's own git root — is registered with the serve hub in ~/.codesearch/repos.json, or a CODESEARCH_SERVER env var is set for remote-serve setups), with a message telling the model how to load and call codesearch instead. Grep is auto-allowed only when the serve hub is genuinely down, established by a live probe of the unauthenticated /healthz endpoint (CODESEARCH_SERVER > 127.0.0.1:$CODESEARCH_SERVE_PORT > 127.0.0.1:39725); only a connection-level failure counts as down. A low-confidence or empty codesearch result is a successful call meaning "reformulate the query", so it does not open the escape hatch — the deny message steers to find/explore/a single clean term instead. Greps outside any registered repo are never blocked, and the hook fails open (never traps the model).edit-guard — on Edit/Write/MultiEdit. Blocks an edit to a file in a codesearch-registered repo until codesearch was consulted for that exact path within the last 5 minutes — find_impact for SCIP-backed languages (.cs .ts .tsx .mts .cts), find(kind="usages") for everything else — making the caller-aware-editing protocol structural. Its companion edit-guard-post (PostToolUse) records the markers on every find_impact call and every find(kind="usages"); any outcome counts ("no results" included), so the guard can never wedge. Unregistered repos, non-git paths and hook failures fail open.subagent-preamble — on Agent (the subagent-spawn tool). Prepends a short codesearch preamble to every subagent prompt, since subagents otherwise don't inherit AGENTS.md or MCP instructions at all — including the edit-guard protocol above.web-guard — on WebSearch/WebFetch. When you have remote documentation projects mounted (codesearch remote mount, e.g. cloud/inriver, cloud/example-dam), it blocks the first web call with guidance to search those indexed mounts first — often more precise and current than the open web. Same 5-minute retry-escape; when no mounts are configured it does nothing.Install (idempotent — user scope applies to every project; --project is this repo only):
codesearch hooks claude install # preferred — self-contained, all platforms
codesearch hooks claude install --project # project scope (./.claude)
The native command embeds the hook scripts in the binary (no source tree needed) and merges the registrations into settings.json. The equivalent from-source installers still live in integrations/claude-code/ (install.ps1 / install.sh) if you'd rather run them directly.
OpenCode gets the same structural treatment through the OpenCode v2 plugin in integrations/opencode/: it resolves the session directory to an indexed project, injects scope-aware guidance once per session, rescues empty greps with index hits, optionally prunes or blocks grep/glob, keeps the serve hub healthy (with optional auto-start), and adds /codesearch* commands, a codesearch_scope tool, a skill and compaction assistance. Install by copying codesearch.ts into ~/.config/opencode/plugins/ — see the integration README.
Note: the grep-guard detects "codesearch is available for this repo" via that repo's registration with the serve hub (~/.codesearch/repos.json, honoring the CODESEARCH_REPOS_CONFIG / CODESEARCH_HOME overrides) or CODESEARCH_SERVER — not by checking whether a codesearch process is running (that runs almost constantly as a multi-repo hub and would false-fire in every directory), and not via a local .codesearch.db directory (a stale db from a since-unregistered repo used to deny Grep even though the hub could not answer for it). For a remote-serve setup with no local registration, set CODESEARCH_SERVER to opt back into enforcement.
search — Code Search
filter_pathsemantics by routing mode:
- Local project (stdio, or
project=<local-alias>/local group on aservehub):filter_pathis a repo-relative prefix (e.g.src,docs/api) — matched against the routed project's own root, not the<alias>/…prefix shown in results.- Federated / mounted project (
project=<peer>/<alias>or an@peergroup fan-out):filter_pathis matched client-side on the namespaced result path (<peer>/<alias>/…) — i.e. exactly the path you see in the results — because the peer only matches its own un-namespaced store paths. The hub over-fetches from the peer and post-filters.Both modes now work without any client-side workaround; earlier releases dropped every hit when
filter_pathwas combined with a serve-routed or federated project.
Semantic mode combines vector similarity (fastembed) + BM25 lexical scoring + exact identifier boosting, fused with RRF. Best for conceptual queries and mixed natural-language + symbol searches.
Literal mode uses Tantivy FTS. Use regex=true for patterns with punctuation (foo::bar, Vec<T>). Use phrase=true for multi-word exact matches.
find — Symbol Navigationexplore — File ExplorationOutline returns all top-level symbols in a file (kind, signature, line range). Similar finds semantically related chunks to a given chunk_id.
get_chunk — Read CodeIn multi-repo mode: auto-routes when chunk_id is unique; returns candidates list when ambiguous.
find_impact — Symbol Reference ImpactFind all call-sites and references to a symbol with file/line precision — the recommended tool for "who calls X?" / "what breaks if I rename X?". Powered by per-language SCIP semantic analysis; precision backends ship per language: C# (bundled scip-csharp helper) and TypeScript (via npx scip-typescript, resolved on the host on demand — no bundle shipped), more planned. When no backend is available for a language, find_impact reports it — fall back to find kind="usages" (lexical) only then.
Returns a list of references with file, start_line, end_line, and kind (e.g. "call", "definition"). Exposes index_age_seconds so agents can reason about staleness.
Note: SCIP precision requires the
-with-csharprelease variant (or a separately installedscip-csharphelper) for C#, andnpx(withscip-typescript, fetched on first use) on the host's PATH for TypeScript. Without a backend for a language,find_impactreturns a clear message — usefind kind="usages"as the lexical fallback. See C# Semantic Search.
On a running codesearch serve the same lookup is available over plain REST — POST /find-impact with the identical request body, no MCP session required (same auth class as the other read-only REST endpoints).
status — Index InfoFor working across multiple repositories simultaneously:
codesearch serve
This starts a background HTTP server with:
Repos are registered via codesearch index add:
# Register a repo (creates index + adds to ~/.codesearch/repos.json)
codesearch index add /path/to/my-project
# Remove a repo
codesearch index rm /path/to/my-project
# List registered repos
codesearch index list
# Clean up stale entries (relocates moved repos, drops the rest)
codesearch index prune
The repository alias (the key in repos.json, used for groups and the MCP
project argument) is always derived automatically from the directory name —
there is no --alias flag.
Serve reads ~/.codesearch/repos.json on startup and manages all registered repos.
If you rename or move a registered folder, serve does not crash. On startup
it tries to relocate each missing repo automatically: it captures every
repo's git remote (remote.origin.url) at registration, and on a missing path
it scans nearby folders (bounded depth, override with
CODESEARCH_RELOCATE_MAX_DEPTH, default 3) for a git checkout with the same
remote. A single unambiguous match is rewritten into repos.json; otherwise the
entry is logged and skipped (never indexed against a dead path). Run
codesearch index prune to relocate what can be relocated and drop the rest.
A hand-edited repos.json is also tolerated: empty entries, orphaned metadata,
and group references to unknown repos are cleaned up on load rather than
crashing.
Groups let you search across related repositories:
codesearch groups add my-group --aliases repo1 repo2 repo3
codesearch groups list
Then in MCP tools: group="my-group" fans out the query to all repos in the group.
all groupThe name all is a reserved virtual group that always resolves to every registered repository — no setup required:
group="all" # fans out to all repos, equivalent to listing every alias
repos.json and always reflects the current set of registered repos (register or remove a repo and all updates automatically).codesearch groups list (marked virtual) and in the scope_required error's available_groups, so agents discover it without extra setup.project/group is specified in multi-repo mode, codesearch still returns scope_required (safe-by-default). Use group="all" explicitly when you want to search everywhere.codesearch groups add all and codesearch groups remove all are rejected — the name is reserved.When using git worktree add to create parallel working directories, codesearch can auto-register new worktrees via a post-checkout git hook.
Setup (run inside any repo you want worktree auto-indexing for):
codesearch hooks git install
This installs a post-checkout hook that POSTs the worktree path to the running serve instance whenever a new worktree is checked out. The hook reads the serve URL from ~/.codesearch/serve_url ($CODESEARCH_HOME/serve_url when the global root is relocated; automatically managed by codesearch serve).
The install target is resolved with git rev-parse --git-path hooks, so it honours core.hooksPath (and, inside a linked worktree, the shared common-dir hooks) rather than assuming .git/hooks/. An existing post-checkout is not overwritten — codesearch's logic is chained in as a marker-delimited block.
How it works:
codesearch serve writes its URL to ~/.codesearch/serve_url ($CODESEARCH_HOME/serve_url when relocated) on startup (deletes on shutdown)post-checkout hook reads that file and POSTs the working directory to POST /reposcodesearch hooks claude install (--project for repo scope) installs the four PreToolUse guard hooks that steer agents to codesearch before Grep/WebSearch/WebFetch/Edit/Write/MultiEdit, plus the PostToolUse marker hook (edit-guard-post) that records the consultations edit-guard requires. See Agent Guidance above for what each guard does.
The codesearch mcp command supports three modes:
codesearch mcp --mode client # force serve connection
The serve endpoint is available at /mcp (Streamable HTTP transport).
codesearch can fan-out read queries (search, get_chunk) to remote peers — other codesearch serve instances (e.g. a cloud-hosted docs/KB peer) — and merge results with local indexes via RRF. This lets a team share one knowledge base while each dev keeps code search local.
Manage peer entries (pure local config — does not call the remote):
codesearch remote add cloud \
--url https://codesearch-serve.<env>.<region>.azurecontainerapps.io \
--api-key $API_KEY --timeout-secs 90
codesearch remote list # show configured peers
codesearch remote rm cloud # remove a peer entry
A group then references a peer via @-prefix ("groups": { "docs": ["@cloud"] }), and group="docs" fans the query out over TLS. Remote misses never hard-fail — they degrade to local-only results with a warnings field.
Manage indexes ON a peer — the same index verbs, scoped with --remote <peer>:
# list the repos living on the cloud peer
codesearch index list --remote cloud
# register a path on the peer's filesystem (NOT your local FS)
codesearch index add /data/docs/vendor-docs --remote cloud
# remove a repo by its alias on the peer (NOT a local path)
codesearch index rm inriver --remote cloud
# trigger a background reindex of one repo on the peer
codesearch index reindex inriver --remote cloud # incremental
codesearch index reindex inriver --remote cloud --force # force full
--remote resolves <peer> against the peers you configured with codesearch remote add. An unknown peer produces a clear error listing the known ones.--remote, add takes a path on the peer's filesystem and rm/reindex take a remote alias (never your local path).--remote, every index command behaves exactly as today (local).index list and index reindex accept --json for agent-friendly output (requires --remote).add, reindex, --force) require the peer to hold the repo read-write. A read-only / restore-only peer (e.g. a snapshot-restore cloud serve) rejects them — --force returns a clean HTTP 500 with a message, and add (a full embed) may OOM or time out a small replica. Use them against a writable peer; list is always safe.Per-vendor layout on a peer. Instead of registering one mixed corpus, register each vendor's sub-folder as its own repo so the peer's layout mirrors your local one. (Requires a writable peer — see the note above; a read-only restore-only peer rejects add.)
for v in vendor-a-docs vendor-b-docs vendor-c-docs; do
codesearch index add "/data/docs/$v" --remote cloud
done
codesearch index list --remote cloud # one alias per vendor
Mounting a peer's projects (opt-in, query one by name). Adding a peer does not expose its projects automatically — you pick the individual indexes you want. Inspect what a peer offers, then mount the ones you care about:
codesearch remote available cloud # list the peer's projects, ✓ marks mounted
codesearch remote mount cloud/akeneo # opt in to a single index
codesearch remote mounts # show what you've mounted
codesearch remote unmount cloud/akeneo # opt back out
A mounted project is addressable directly as project=<peer>/<alias> — a 1-to-1 passthrough to that peer's index:
# search only the peer's akeneo docs, by name
codesearch search "import products" --project cloud/akeneo
The remote_mounts allowlist in ~/.codesearch/repos.json is the single source of truth: a non-mounted project is unroutable (even if the peer exposes it), and a whole-peer @peer group reference (e.g. docs → [@cloud]) federates only your mounted indexes for that peer, not its whole corpus. Mounted projects are surfaced by list_projects (a remote_projects array) and advertised in the scope_required error, so agents can discover them.
In the codesearch serve TUI, mounts appear in italic/cyan, distinguishing them from local indexes. Press i on a mount to see its Remote Mount info (peer URL + peer-reported status); the local-index actions (doctor / reindex / remove) are shown struck-through/disabled for a mount, since they act on a local index and a mount has none — manage a peer's indexes with --remote (above) instead.
A serve hub with many repos used to start one full refresh/reindex/HNSW-build
pass per repo as soon as that repo had work. A mass git pull, a burst of
first queries, or several POST /repos/:alias/reindex calls therefore ran
many CPU-heavy passes at once, held many file/chunk windows in memory, and
queued the (already serialised) ONNX inference behind all of them. These knobs
bound that work; they are read at startup, so a shell or systemd unit can set
them without editing anything:
codesearch serve also accepts --index-jobs, --embed-threads,
--embed-pause-ms and --min-free-mb, which set the corresponding variable
for that process (flags win over the environment). Standalone codesearch index and codesearch mcp runs have no job gate and keep their previous
behaviour; the embedding and batching variables still apply to them.
Example for a large shared machine (128 cores, ~30 repos, dozens of repos indexed at once are the failure mode):
# ~/.config/systemd/user/codesearch.service.d/limits.conf
[Service]
Environment=CODESEARCH_INDEX_JOBS=1
Environment=CODESEARCH_EMBED_THREADS=8
Environment=CODESEARCH_INCREMENTAL_BATCH_SIZE=100
Environment=CODESEARCH_MAX_CHUNKS_PER_BATCH=5000
Environment=CODESEARCH_EMBED_PAUSE_MS=25
Environment=CODESEARCH_MIN_FREE_MB=4096
# C# SCIP helpers are bounded separately; lower it for the same reason.
Environment=CSHARP_SCIP_CONCURRENCY=1
.codesearchignorePlace in repo root. Gitignore syntax. Excludes paths from indexing:
# Vendored code
vendor/
node_modules/
# Generated files
*.generated.cs
**/migrations/**
A global .codesearchignore can be placed at ~/.codesearch/.codesearchignore. It applies to all repos with the lowest priority (repo-local .codesearchignore, .gitignore, and .git/info/exclude all override it). This is useful for patterns you want everywhere without modifying each repo.
By default codesearch recognises the fixed extension list in Supported
Languages; any other extension is Unknown and is
skipped entirely (never indexed). To teach codesearch about a non-standard
extension — or to deliberately remap a known one — drop a small JSON object at
~/.codesearch/extensions.json mapping extension → language name:
{
"inc": "php",
"phtml": "php",
"h": "cpp"
}
"inc", ".inc", ".INC" are equivalent). Only the last dot-suffix is
matched, so Foo.class.inc maps via "inc".php, cpp/c++, csharp/c#, golang, js, ts, …), case-insensitive."h": "cpp" overrides the default C mapping).CODESEARCH_EXTENSION_MAP to load the file from a different path.This is the supported answer to "my PHP is in
*.incfiles" (issue #138):.incis intentionally not a built-in because it's language-agnostic (assembly, SQL, C/PHP includes all use it), so you opt in per your codebase.
repos.jsonLocated at ~/.codesearch/repos.json. Managed by codesearch index add/rm. Contains repo aliases → paths and group definitions. See Serve Mode.
This section documents the security model of federation / remote peers — the largest network surface in codesearch — followed by serve access control.
When codesearch serve is exposed beyond a single trusted user (shared dev machines, a network bind), two environment variables harden access:
CODESEARCH_SERVE_API_KEY — gates access depending on how serve is bound:
--host 0.0.0.0, a LAN IP, etc.) — ALL endpoints require the key (health, status, MCP search, and management). Setting this variable is required when binding to a non-localhost address; serve refuses to start without it.POST /repos, DELETE /repos/:alias, POST /repos/:alias/reindex, POST /reload) require the key. Health, status, and MCP search remain open.Authorization: Bearer <key> or X-API-Key: <key>.codesearch index add/rm/reindex delegate to a running serve, so the CLI must send the same key. Set CODESEARCH_SERVE_API_KEY in the client's environment (same value as the server) and the CLI attaches it automatically. Without it, delegation returns 401 and falls back to local indexing — which risks LMDB file-lock conflicts if serve is still running.CODESEARCH_ALLOWED_ROOTS — semicolon-separated list of filesystem roots. Repo registration is rejected for paths outside these roots. Prevents indexing arbitrary directories.Both are backward compatible: unset means no restriction (on a localhost bind).
The MCP streamable-HTTP transport (via rmcp) validates the incoming Host header against an allowlist to defend against DNS-rebinding attacks. By default this allowlist is loopback-only (localhost, 127.0.0.1, ::1), which rejects requests carrying a container hostname or service-discovery name — a common trip-up in containerised/orchestrated deployments (Docker, Kubernetes, etc.) where the client connects via a non-loopback Host header.
CODESEARCH_ALLOWED_HOSTS — comma-separated list of extra allowed hostnames (e.g. codesearch-serve,codesearch-serve.internal), replacing the loopback-only default. Prefer this over disabling validation.CODESEARCH_DISABLE_HOST_VALIDATION — set to 1 or true to disable Host-header validation entirely. This removes the DNS-rebinding protection outright; only use it when serve is already fenced off by a reverse proxy or network policy that restricts which Host headers can reach it.Precedence: disable > custom allowlist > default (loopback-only).
Beyond the access-control gates above, codesearch applies several defense-in-depth mitigations at the filesystem and CLI boundary:
index, repo registration) fails fast on an unresolvable/malformed path instead of silently falling back to the raw, unvalidated input..NET symbol-helper CLI (scip-csharp) canonicalizes every path argument (--solution, --project, --output, --symbols-file) before use..git, .svn, node_modules, etc.) is rejected, preventing accidental indexing/exposure of internal VCS metadata.codesearch serve's file-descriptor demand scales with the number of registered repos (each warm repo holds LMDB + full-text-index + file-watcher handles). Under a process supervisor with a low default open-file limit (notably macOS launchd, default soft ulimit -n 256), a large repo count can silently exhaust file descriptors: accept() then fails with EMFILE and the daemon looks alive to its supervisor while refusing new connections. Serve now raises its own soft RLIMIT_NOFILE to the hard limit at startup (Unix only) and logs a warning if the effective limit still looks insufficient for the registered repo count — but if you see repeated EMFILE/"Too many open files" in the logs, raise the hard limit for the service (e.g. launchd SoftResourceLimits/HardResourceLimits, systemd LimitNOFILE=, or ulimit -n in the service's environment).
Federation is operator-to-operator, not end-user-facing. The only inputs that decide where requests go and which key they carry are the peer entries you register locally with codesearch remote add (stored in ~/.codesearch/repos.json). No search query, MCP argument, or remote response ever becomes a request target or selects a key.
api_key is stored in plaintext in ~/.codesearch/repos.json (your home directory, outside any git repo — it is never committed). Protect it with filesystem permissions, the same way you would an SSH key or a .env file. On the server side, inject the key from a secret store — the reference cloud deployment sets it as an Azure Container Apps secret (secretref:api-key), so it never sits in the image or in source.Authorization: Bearer <key> header, over HTTPS. Use https:// peer URLs; the federation client uses rustls with certificate verification enabled — there is no certificate-bypass (danger_accept_invalid_certs) anywhere in the codebase.Authorization header of the in-flight request.The HTTP client follows up to 10 redirects by default (reqwest's standard policy) but strips the Authorization header on cross-host redirects. A peer that answers with a 3xx to a different host therefore cannot exfiltrate the bearer token there; same-host redirects (e.g. path canonicalisation) preserve it. Defense-in-depth: if you suspect a peer's host or DNS has been compromised, rotate that peer's key.
The remote serve instance enforces its own auth on every inbound request from federation:
/search and /chunk/:id — require the key. The reference cloud peer binds 0.0.0.0, so every federation fan-out carries the key and is rejected without it.POST /repos, DELETE /repos/:alias, POST .../reindex, POST /reload) require the key on every bind, localhost included.The @peer fan-out attaches the configured key via the same bearer_auth path the local CLI uses, so it satisfies both layers automatically.
index … --remote)add, rm, and reindex against a peer are authenticated management calls. They carry the key like any other request and the peer decides whether to act (a read-only / restore-only peer rejects writes — see Federation). The local CLI only sends a path or alias for the peer to act on; it executes nothing on your machine and writes nothing to your filesystem.
On fan-out, the client strips the local project and forces group to the value configured for that peer (or all). Projects are local to each instance, so this prevents one instance's project names from leaking into another's query namespace.
All C#-specific setup, operation, installation, and testing lives in README_CSharp.md.
If you do not work with C# repos, you can skip it entirely.
Tree-sitter AST-aware chunking:
Markdown uses the tree-sitter-md block grammar — chunks align to sections,
headings, and code fences. Jupyter notebooks are parsed as JSON; code and
markdown cells are extracted, tagged with [code] or [markdown], and
adjacent same-type cells under 50 lines are merged into single chunks.
All other text files use line-based chunking as fallback.
Files whose extension isn't recognised are treated as Unknown and skipped
entirely (not indexed). If your codebase uses a non-standard extension for a
supported language — e.g. legacy PHP in *.inc files, or .h you want parsed
as C++ — map it explicitly with the extension map.
# Build
cargo build
# Run tests
cargo test
# Check + lint
cargo clippy --all-targets -- -D warnings
# Format
cargo fmt --all
Apache-2.0
Built with: fastembed-rs, arroy, tantivy, tree-sitter, ratatui, LMDB.
Multi-repo semantic code search MCP server in Rust — hybrid vector + BM25 retrieval, tree-sitter AST chunking, fully offline. For OpenCode, Claude Code, Cursor, and any MCP client.
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式，而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Codesearch 透過 stdio 在本機執行。
上次檢查時間：2026-11-11。 Multi-repo semantic code search MCP server in Rust — hybrid vector + BM25 retrieval, tree-sitter AST chunking, fully offline. For OpenCode, Claude Code, Cursor, and any MCP client.