Examples

Table of Contents

Basic Stdio Configuration

Configure a Chroma collection populated by chromadb-repo-indexer and an embedding endpoint compatible with the collection:

export CHROMA_REPO_SEARCH_SERVER_URL=https://chroma.example.com
export CHROMA_REPO_SEARCH_COLLECTION_NAME=repository-content
export CHROMA_REPO_SEARCH_BEARER_TOKEN=your-chroma-token
export CHROMA_REPO_SEARCH_EMBEDDING_API_URL=https://embeddings.example.com
export CHROMA_REPO_SEARCH_EMBEDDING_MODEL=your-embedding-model
export CHROMA_REPO_SEARCH_EMBEDDING_API_KEY=your-embedding-key

mcp-chromadb-repo-search

Default Chroma tenant and database values are default_tenant and default_database.

Shared Indexer Configuration

The search server can read relevant connection and embedding values from the same version 1 YAML file used by the indexer:

version: 1

chroma:
  server_url: https://chroma.example.com
  collection_name: repository-content
  tenant: default_tenant
  database: default_database

files:
  include_paths:
    - "**"
  exclude_paths:
    - "vendor/**"

chunking:
  chunk_size: 512
  chunk_overlap: 64

sync:
  batch_size: 100
  retry_attempts: 3

embedding:
  api_url: https://embeddings.example.com
  model: your-embedding-model
  api_key: ""

export CHROMA_REPO_SEARCH_BEARER_TOKEN=your-chroma-token
mcp-chromadb-repo-search --config /path/to/config.yml

File selection and chunking fields are used by the indexer and safely accepted by the search server, but they do not alter query behavior.

Search Every Indexed Repository

Omit source when the relevant repository is unknown:

{
  "query": "Where is exponential retry backoff implemented?"
}

Search One Repository 

{
  "query": "How is the application configuration validated?",
  "source": "UnitVectorY-Labs/mcp-chromadb-repo-search",
  "limit": 5
}

This searches all indexed branches for that repository.

Search One Branch 

{
  "query": "How does the MCP server select its transport?",
  "source": "UnitVectorY-Labs/mcp-chromadb-repo-search@main"
}

Filter by Path

Search Go files anywhere in the repository:

{
  "query": "How are duplicate search results removed?",
  "source": "UnitVectorY-Labs/mcp-chromadb-repo-search@main",
  "path": "**/*.go",
  "limit": 3
}

Search documentation under docs:

{
  "query": "What environment variables configure reranking?",
  "source": "UnitVectorY-Labs/mcp-chromadb-repo-search@main",
  "path": "docs/*"
}

Search one exact file:

{
  "query": "What is this project for?",
  "path": "README.md"
}

Enable Reranking 

export CHROMA_REPO_SEARCH_RERANK_API_URL=https://rerank.example.com
export CHROMA_REPO_SEARCH_RERANK_MODEL=your-reranker-model
export CHROMA_REPO_SEARCH_RERANK_API_KEY=your-reranking-key
export CHROMA_REPO_SEARCH_RERANK_CANDIDATE_MULTIPLIER=3
export CHROMA_REPO_SEARCH_RERANK_MAX_CANDIDATES=100

mcp-chromadb-repo-search

For a search with limit: 5, the default settings retrieve up to 15 Chroma candidates, rerank them, and return up to five unique, file-diverse excerpts.

Streamable HTTP

Start the server on the loopback interface:

mcp-chromadb-repo-search --http 127.0.0.1:8080

Configure a remote-capable MCP client to use:

http://127.0.0.1:8080/mcp

If no static Chroma bearer token is configured, an Authorization header received at this endpoint is forwarded to Chroma. The endpoint itself must be protected separately for non-local deployments.

Direct MCP Testing

List Tools over Stdio

The stdio transport accepts newline-delimited JSON-RPC messages. Listing tools validates server startup without making a backend request:

printf '%s\n' \
  '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"manual-test","version":"1"}}}' \
  '{"jsonrpc":"2.0","method":"notifications/initialized"}' \
  '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' \
  | mcp-chromadb-repo-search \
      --server-url https://chroma.example.com \
      --collection-name repository-content \
      --embedding-api-url https://embeddings.example.com \
      --embedding-model your-embedding-model

Call search over Stdio

Add this JSON-RPC message after initialization to perform a backend search:

{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"search","arguments":{"query":"how is configuration loaded?","source":"example-org/example-repository@main","path":"internal/config/*","limit":3}}}

Test Streamable HTTP

Send an MCP initialize request to /mcp, then call tools/call according to the MCP Streamable HTTP protocol. This server is stateless, so it does not require a server-maintained MCP session between requests.

Example Result 

Found 1 relevant repository excerpt, ranked by semantic relevance.

---

### 1

**Repo:** `example-org/example-repository@main`
**File:** `internal/config/config.go#L20-L32`

Source: [internal/config/config.go](https://github.com/example-org/example-repository/blob/COMMIT/internal/config/config.go#L20-L32)
Context: symbol `Load`

```go
func Load() error {
    // Relevant source excerpt
}
```

The source link uses the indexed commit SHA when available, making the result reproducible even after the branch advances.