MCP代码图引擎

1 · install the native runtime (signed, verified, with rollback)

npx -y @maxkle1nz/m1nd update apply --yes

3 · wire your host: MCP config + the session hooks that make m1nd ambient

npx -y @maxkle1nz/m1nd hosts apply --host claude --project . --yes

5 · first value: the orientation packet for YOUR repo, read-only, no host config touched

npx -y @maxkle1nz/m1nd agent first-minute --repo . --query "map this repo" --json ```

Step 4 is the one command only you run. Minting a brain writes a whole graph, so the ceremony has a human terminal ingress and accepts only an empty destination. If an agent finds a repo without a brain, it offers this command and stops. Once the graph exists, the agent can keep it fresh. The ceremony exits non-zero and tells you what to check if the scan finds nothing, so it can never report success over an empty graph.

Step 1 verifies the signature with cosign, so install that first if it is not on your PATH. If you prefer the source registry and accept skipping verification, cargo install m1nd-mcp works too. Prefer to see before you write: hosts plan prints everything hosts apply would touch, and writes nothing. There is no uninstall command yet; hosts plan doubles as the list of what to remove by hand.

The hooks from step 3 are what make m1nd ambient: the orientation packet is injected at every session and subagent spawn, and the agent drives itself from there. Installing from an agent instead of a terminal? There is a machine-legible twin of this section in llms-install.md.

A tampered or truncated release cannot land on your machine, and a bad upgrade is one rollback away: the updater checks the signature against the exact build identity, then the SHA-256 and the size, before it touches anything. If verification fails, it refuses rather than falling back to an unverified path. Details in docs/AGENT-PACKS.md.

grep answers good questions. m1nd answers the deeper ones.

Questions your agent can now ask and get a structural answer for:

• What breaks if I touch this function?
• Where does token refresh actually happen in this repo?
• Why are these two files connected, and is that path solid or a guess?
• What did the last session learn about this code, and is it still true?
• What always changes together here, even with no import between them?
• Does this edit cross an architecture boundary I should not cross?
• Which claim in this paper does this function implement?
• Is the bug I just fixed hiding anywhere else, as a shape?
• What is missing here that this pattern usually has?
• Am I even in the right repo?
• Should I act on this answer, or verify it first?

Each one is a verb on the MCP surface (impact, seek, why, north, ghost_edges, xray_gate, antibody_scan, missing, trust_selftest, predict), not a prompt trick.