MCP server that wraps the Claude Code CLI as a subprocess and exposes Claude capabilities as MCP tools for query, code review, web search, structured JSON output, health checks, and session listing. It supports session resume, effort and budget controls, and returns cost metadata in responses.
From the registry: Wraps Claude Code CLI as MCP tools: query, search, structured, ping, listSessions.
$ claude loginPlease install the `claude-mcp-bridge` MCP server into my current AI client (that's you).
Required prerequisites (do these first if not already done):
- **Install Node.js** — Node.js is required to run the server via npx. (https://nodejs.org/)
- **Install Claude Code CLI** — Install the Claude Code CLI and ensure the `claude` command is available on PATH. (https://github.com/anthropics/claude-code)
- **Authenticate Claude Code CLI** — Use one authentication method: either log in with your Claude subscription (default), or configure API-key auth by setting `ANTHROPIC_API_KEY` and `CLAUDE_BRIDGE_USE_API_KEY=1`. Run: `claude login`
Canonical MCP server config (stdio transport):
- command: `npx`
- args: ["-y","claude-mcp-bridge"]
- optional environment variables:
- `ANTHROPIC_API_KEY`: Anthropic API key. Required only if using API-key auth instead of subscription auth. (example: `<your-anthropic-api-key>`)
- `CLAUDE_BRIDGE_USE_API_KEY`: Set to `1` to forward `ANTHROPIC_API_KEY` to the Claude CLI subprocess and use API-key billing instead of subscription auth. (example: `1`)
- `CLAUDE_DEFAULT_MODEL`: Shared default model for all tools. (example: `<your-default-model>`)
- `CLAUDE_QUERY_MODEL`: Default model for the `query` tool. (example: `<your-query-model>`)
- `CLAUDE_STRUCTURED_MODEL`: Default model for the `structured` tool. (example: `<your-structured-model>`)
- `CLAUDE_SEARCH_MODEL`: Default model for the `search` tool. (example: `<your-search-model>`)
- `CLAUDE_REVIEW_MODEL`: Default model for the `review` tool. (example: `<your-review-model>`)
- `CLAUDE_FALLBACK_MODEL`: Fallback model to use on quota exhaustion; set to `none` to disable fallback. (example: `<your-fallback-model>`)
- `CLAUDE_MAX_CONCURRENT`: Maximum number of concurrent Claude subprocesses. (example: `<max-concurrent>`)
- `CLAUDE_CLI_PATH`: Path to the Claude CLI binary if not available as `claude` on PATH. (example: `<path-to-claude-cli>`)
- `CLAUDE_MAX_BUDGET_USD`: Global per-call cost cap in USD. (example: `<max-budget-usd>`)
- `CLAUDE_REVIEW_EFFORT`: Default effort level for the `review` tool. (example: `<review-effort>`)
- `CLAUDE_SEARCH_EFFORT`: Default effort level for the `search` tool. (example: `<search-effort>`)
- `CLAUDE_QUERY_EFFORT`: Default effort level for the `query` tool. (example: `<query-effort>`)
Note: The preferred install path is `npx -y claude-mcp-bridge`. This is a stdio MCP server that wraps the local Claude Code CLI. By default it uses Claude subscription auth via `claude login`; API-key auth is optional and requires both `ANTHROPIC_API_KEY` and `CLAUDE_BRIDGE_USE_API_KEY=1`.
Add this MCP server to my current client's config in the correct format for you. If you need secrets or credentials I haven't provided, ASK me — do not invent values or leave raw placeholders. After adding it, tell me how to verify the server is connected.ANTHROPIC_API_KEYrequiredAPI key used when opting into API-key authentication for the Claude CLI subprocess.CLAUDE_BRIDGE_USE_API_KEYrequiredSet to 1 to forward ANTHROPIC_API_KEY to the subprocess instead of using subscription authentication.AI orchestration with hive-mind swarms, neural networks, and 87 MCP tools for enterprise dev.
Persistent codebase knowledge graph. Survives session restarts and context compaction.
A powerful toolkit for coding, providing semantic retrieval and editing capabilities.