HAI-MCP server

Read from the package source without running it. The installed server may list more.

• hai_accept_next_step

  Promote proposed (or provided) content to canonical NEXT_STEP.md — owner-gated.

• hai_authorize_session

  Grant a time-bounded session lease tied to mission ID and exact contract version.

• hai_bind_project

  Bind a device-local directory to a logical project_id mount table entry. Requires owner_ack=true + reason.

• hai_check_activity

  Deterministically classify planned or observed activity against the mission contract.

• hai_checkpoint

  Snapshot ACTIVE_CONTEXT and optional project artifacts under HAI_HOME/history/checkpoints.

• hai_close_mission

  Complete with verified per-criterion evidence, or abandon (owner-gated).

• hai_distill

  Reduce an intake to EXACTLY one decision + one next step; the server parks everything else.

• hai_drift_check

  Lightweight mismatch check against the mission contract. Thin wrapper over hai_check_activity.

• hai_get_contract

  Return the exact canonical mission contract for a valid session lease (not a summary).

• hai_get_next_step

  Read the canonical NEXT_STEP.md for a project (or report none).

• hai_health

  Check HAI-MCP server health and optional project path usability. No model calls.

• hai_intake

  Capture a raw thought immutably. Returns an intake id only — never actionable, never starts an agent.

• hai_mission_start

  Fast start: declare problem + artifact + time limit. Thin wrapper over hai_open_mission (one canonical contract).

• hai_open_mission

  Open a bounded mission with a versioned canonical contract. One active mission globally.

• hai_park

  Park a thought/meta item in HAI inbox without changing ACTIVE lanes.

• hai_park_item

  Park a mission-linked idea with full context. Grants no execution right; contract unchanged.

• hai_proof

  Close a mission only against verified per-criterion evidence. Thin wrapper over hai_close_mission (completed).

• hai_propose_next_stepwrite action

  Write NEXT_STEP.proposed.md (not canonical). Promote via hai_accept_next_step.

• hai_read_artifacts

  Read-only summary of HAI Projek-Managment run-contract artifacts.

• hai_recontract

  Apply a visible field-level contract diff; revokes all leases. Owner-gated.

• hai_recover

  Return the smallest recovery next action from latest or named checkpoint (read-only advice).

• hai_set_focus

  Set focus and register an ACTIVE lane (max 2). Fails if a third ACTIVE is requested.

• hai_status

  Return ACTIVE lanes, focus, inbox count, and optional project next-step flags.

• hai_stop

  Hard day terminal: record the three answers, invalidate active leases; no next-day plan. Missions are not closed.

• Code scan19 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitystatic API keys via environment variables6/15
• Maintenancelast push 0 days ago15/15
• Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Safety score

83/100

higher than 68% of 4,460 AI and agents servers

Tools

24

the typical AI and agents server has 6

Tools that change things

4%

typical: 0%. Fewer write tools means less an agent can break

HAI-MCP: common questions

Is HAI-MCP server safe?

Mostly: it is graded B (83/100). Read the HAI-MCP safety report

How do I install HAI-MCP?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does HAI-MCP need an API key?

Yes. The registry entry asks for HAI_HTTP_TOKEN.

Is HAI-MCP maintained?

The last commit was 22 days ago (2026-10-13). The latest release is v0.1.0.

What can I use instead of HAI-MCP?

Servers from other publishers that do the same job: DCC-MCP Core MCP server, agent-device MCP server and agent-device MCP server.

How does HAI-MCP compare with other AI and agents MCP servers?

safety score 83/100 (higher than 68% of 4,460 AI and agents servers); tools 24 (the typical AI and agents server has 6); tools that change things 4% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.