Route research queries to Gemini CLI through a stateless MCP server, reducing agent context overhead and model token usage. For developers building AI agents that need lightweight, cached research capabilities without bloating conversation history.
io.github.capyBearista/gemini-researcher
Local install
STDIO
No auth required
How models use it and what it is built for.
Route research queries to Gemini CLI through a stateless MCP server, reducing agent context overhead and model token usage. For developers building AI agents that need lightweight, cached research capabilities without bloating conversation history.
Local install — runs as a subprocess.
Configuration this server reads at startup.
Gemini API key (optional if you already authenticated Gemini CLI via "gemini" login)
Override the project root directory used for path validation (defaults to current working directory)
Chunk size threshold (KB) for large responses (default: 10)
Chunk cache TTL in milliseconds (default: 3600000 / 1 hour)
Enable debug logging (set to "true" or "1")
Vertex AI / Google auth: path to a service account JSON credentials file
Vertex AI / Google auth: GCP project ID (used by some auth configurations)
Vertex AI / Google auth: Vertex AI project identifier (used by some auth configurations)
Where to find authoritative docs and source for Capybearista Gemini Researcher.
Paste any of these into Agent Studio after connecting Capybearista Gemini Researcher.
Common questions about connecting and running Capybearista Gemini Researcher.
How do I authenticate this MCP server with Gemini?
Set GEMINI_API_KEY with your Gemini API key, or authenticate via the Gemini CLI using 'gemini login' first. For Vertex AI, provide GOOGLE_APPLICATION_CREDENTIALS (path to service account JSON) and GOOGLE_CLOUD_PROJECT or VERTEX_AI_PROJECT.
Does this server cache research results?
Yes. Responses are chunked and cached with a default TTL of 1 hour (3600000 ms). You can adjust CACHE_TTL_MS and RESPONSE_CHUNK_SIZE_KB to control caching behavior and chunk thresholds.
Why would I use this instead of calling Gemini directly in my agent?
This server reduces agent context and model token usage by proxying research queries statelessly. It's designed to keep conversation history lean while offloading research work to Gemini CLI.
Can I override the project root directory for path validation?
Yes, set PROJECT_ROOT to override the default (current working directory). This is useful if your MCP server runs in a different directory than your project.
How do I enable debug logging?
Set DEBUG=true or DEBUG=1 to enable debug logging. This helps troubleshoot authentication, caching, and query routing issues.
MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 60+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.
Open Agent StudioJudges Panel
45 judges that evaluate AI-generated code for security, cost, and quality with built-in AST.
Addozhang Nexus
Query Sonatype Nexus Repository (OSS/Pro) for Maven, Python, and Docker artifacts
Cyanheads MCP Ts Template
TypeScript template for building MCP servers with declarative tooling, observability, and auth.
Chernistry Bernstein
Declarative agent orchestration for engineering teams