What it does
The tool sends questions to Exa's neural index in three forms. A plain search returns the pages that best match the meaning of a query, with their text included. A code question returns a context pack of examples and documentation excerpts gathered from across the web, sized to a token budget chosen by the caller. A crawl fetches one known page and returns its full text.
It suits research-style lookups from a terminal or from an AI agent connected over MCP. The MCP server, started with the mcp subcommand, exposes three tools: web_search_exa, get_code_context_exa and crawling_exa. Each accepts an optional result count or token budget. Budgets are applied at roughly four characters per token, and an empty result set is reported as empty.
Running it requires an Exa API key in the EXA_API_KEY environment variable. The key is sent as an x-api-key header and is not printed in output. A container image is also available, and it starts the MCP server by default. CLI exit codes are 0 on success, 1 when an operation fails and 2 for usage errors. The project is released under the MIT licence.
Limitations and requirements
Requires an Exa API key in the EXA_API_KEY environment variable; without it every operation fails with an error naming the variable. Token budgets are clamped to the 1000–50000 range. A page the API cannot fetch, a missing key and a failed API call are reported as tool errors.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.CloudCorpRecords/deepsearch
Neural web search, code context, and page crawling through the Exa API.
v0.1.0As shown on registry.modelcontextprotocol.io when we last read it