Your agent queries csp in natural language (for example "How is authentication handled?") and gets back only the relevant code snippets, without grepping or reading full files.
csp has three complementary setup paths. The recommended setup is using all three, but you can pick and choose based on your needs:
- MCP server: an MCP server for your agent.
- AGENTS.md: an AGENTS.md snippet with instructions for calling
cspvia the CLI. - Sub-agent: a dedicated
csp-searchsub-agent for harnesses that support it.
Add the MCP server
Expose csp as a native tool via MCP so your agent can call it directly. For Claude Code:
claude mcp add csp -s user -- bunx @pleaseai/csp mcpSee MCP server for other harnesses (Cursor, Codex, OpenCode, VS Code, and more).
Install the CLI and add the AGENTS.md snippet
Install the csp CLI, then add the AGENTS.md snippet to your AGENTS.md or CLAUDE.md so the agent knows when and how to call it.
brew install pleaseai/tap/cspInstall the sub-agent
For harnesses that support sub-agents, install a dedicated csp-search sub-agent so search runs in its own context (requires the CLI):
csp init # Claude Code → .claude/agents/csp-search.mdSee Sub-agent for the other harnesses.