Parkour is a content exploration toolkit designed to help LLMs extract high signal, unsummarized web content efficiently. It utilizes clean APIs and a rolling cache to enhance content retrieval and knowledge synthesis.
From the registry: A content exploration toolkit that helps LLMs surface high signal, unsummarized web content.
$ git clone https://github.com/blightbow/parkour-mcp && cd parkour-mcp && uv syncPlease install the `parkour-mcp` MCP server into my current AI client (that's you).
Required prerequisites (do these first if not already done):
- **Clone repository** — Clone and run via uv Run: `git clone https://github.com/blightbow/parkour-mcp && cd parkour-mcp && uv sync` (https://github.com/blightbow/parkour-mcp)
Optional prerequisites:
- Kagi API Key (optional, for search/summarize tools) — Required only for kagi_search and kagi_summarize tools (https://kagi.com/settings?p=api)
Canonical MCP server config (stdio transport):
- command: `uv`
- args: ["--directory","/path/to/parkour-mcp","run","parkour-mcp","--profile","code"]
- optional environment variables:
- `KAGI_API_KEY`: Kagi API key for search and summarize tools (example: `<your-kagi-api-key>`)
- `S2_ACCEPT_TOS`: Set to 1 to enable Semantic Scholar tool (requires accepting TOS) (example: `1`)
- `GITHUB_TOKEN`: GitHub personal access token for higher rate limits (example: `ghp_<your-github-token>`)
Note: 11 tools for web content extraction: kagi_search, web_fetch_sections, web_fetch_exact, web_fetch_js, semantic_scholar, arxiv, github, ietf, packages, discourse, kagi_summarize. Uses frontmatter tool envelopes to steer LLM. Requires cloning. Use --profile desktop for Claude Desktop, --profile code for Claude Code.
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.MCP server for searching Airweave collections with natural language queries.