autopilot-mcp
MCP server for an agent's autopilot: scopes, a daily SOL spend cap, and the propose, execute and undo loop.
Install / Use
claude mcp add nirholas -- npx -y github:nirholas/autopilot-mcpIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
Development & EngineeringSupported Platforms
Tags
Our assessment of autopilot-mcp
autopilot-mcp scores 81/100 on our quality scale, 3429th of 4,698 Development & Engineering skills we index.
Its MCP Server is 13 KB long, well organised into 15 sections with 6 code examples: a thorough specification that gives an agent plenty to work with.
It has 10 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated 6 days ago, so autopilot-mcp is actively maintained.
- Our last check on 2026-11-10 found the source still online.
- It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-11-10. Automated pattern scan on 2026-11-10. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
autopilot-mcp compared with similar skills
All 4 of these similar skills score higher than autopilot-mcp; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| autopilot-mcp (this skill)by nirholas | 81 | 10 | 6d ago | MCP Server |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 112.3k | 8d ago | CLAUDE.md |
| rufloby ruvnet | 100 | 81.9k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 79.8k | today | CLAUDE.md |
Frequently asked questions
- How do I install autopilot-mcp?
- Run
claude mcp add nirholas -- npx -y github:nirholas/autopilot-mcp. The install tabs above show the steps for each supported agent. - Which AI agents does autopilot-mcp work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is autopilot-mcp safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is Apache-2.0-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is autopilot-mcp still maintained?
- The repository was last updated 6 days ago, so autopilot-mcp is actively maintained.
Skill content
View source on GitHubA Model Context Protocol server that gives an AI agent its own execution control plane over stdio. The agent configures the boundaries it is allowed to act within — capability scopes, a daily SOL spend ceiling, auto-execute and confirmation policy — then drives the full propose → review → execute → undo loop. No human in the loop, but every boundary is enforced server-side: an out-of-scope or over-budget action is denied, never executed.
This is the keystone of the autonomous-agent story. Autopilot config and execution used to live only behind the three.ws UI/API; this server exposes them so an agent can manage its own guardrails and act within them.
⚠️ This server is authenticated and write-heavy.
execute_proposalcan move real SOL on Solana mainnet and is irreversible. It sends native SOL only — the agent never sells or sends $THREE (it only buys, holds, or burns it). Scopes, the daily SOL spend cap, and confirmation are all enforced on the three.ws backend — this server cannot bypass them.
Install
npm install @three-ws/autopilot-mcp
Or run with npx (no install):
npx @three-ws/autopilot-mcp
Quick start
Claude Code, one line (supply your three.ws credential):
THREE_WS_API_KEY=sk_live_… claude mcp add autopilot -- npx -y @three-ws/autopilot-mcp
Claude Desktop / Cursor (claude_desktop_config.json or mcp.json):
{
"mcpServers": {
"autopilot": {
"command": "npx",
"args": ["-y", "@three-ws/autopilot-mcp"],
"env": {
"THREE_WS_API_KEY": "sk_live_…"
}
}
}
}
Inspect the surface with the MCP Inspector:
THREE_WS_API_KEY=sk_live_… npx -y @modelcontextprotocol/inspector npx @three-ws/autopilot-mcp
Authentication
Every autopilot endpoint is owner-only, so this server is authenticated. Set THREE_WS_API_KEY to the agent owner's three.ws credential — either a three.ws API key (sk_live_… / sk_test_…) or an OAuth access token. It is sent as Authorization: Bearer … on every request and identifies the owning user; the agent can only manage agents that user owns. The credential can authorize real SOL transfers via execute_proposal — store it like a password. THREE_WS_TOKEN and THREE_WS_BEARER are accepted as aliases.
Tools
| Tool | Type | What it does |
| ------------------------- | ---------------- | ------------------------------------------------------------------------------------------------------------------ |
| get_autopilot_config | read-only | Read the agent's guardrails: enabled, scopes, auto-execute, daily SOL cap, confirmation policy — plus trust. |
| set_autopilot_config | write · idempotent | Update those guardrails with a partial patch (grant scopes, set the daily SOL ceiling, arm/disarm). No action taken. |
| generate_proposals | write | Run the agent's mind: turn high-salience memories + reflections into real, provenance-cited candidate actions. |
| list_proposals | read-only | The proposal queue — ids, kinds, rationales, params, status, cited sources. Filter by status. |
| dryrun_proposal | read-only | Preview what executing a proposal would do (scope / cap / balance checks), without taking any action. |
| adjust_proposal | write · idempotent | Edit a pending proposal's params before executing (tune a threshold, amount, cadence). Validated server-side. |
| execute_proposal | write · destructive | Take the real action. ⚠️ A wallet_transfer moves real SOL on mainnet, irreversible — needs confirm:true. Never sends $THREE. |
| dismiss_proposal | write | Drop a pending proposal; records a feedback memory so the agent stops proposing it. |
| undo_action | write | Reverse a reversible execution (delete the alert rule / briefing); records a feedback memory. SOL transfers can't be undone. |
| list_autopilot_activity | read-only | The append-only signed receipts log — every autonomous action, its outcome, provenance, and signature. |
| compute_trust | read-only | The agent's earned trust level (sandbox → trusted → autonomous), computed from real action history. |
The five read tools reflect live state (proposals, receipts, trust move between calls), so none are idempotent. set_autopilot_config and adjust_proposal are idempotent writes (re-applying the same values is a no-op). generate_proposals, execute_proposal, dismiss_proposal, and undo_action each take a real action.
Input parameters
Every tool takes agentId (the UUID of an agent you own) except list_autopilot_activity, where it is optional (omit to aggregate across all your agents).
get_autopilot_config/compute_trust/generate_proposals—agentId.set_autopilot_config—agentId, plus any of:enabled(bool),scopes({ create_alert?, briefing?, wallet_transfer? }),autoExecute({ create_alert?, briefing? }),dailySpendSol(number, SOL; 0 disables spending),requireConfirm(bool).list_proposals—agentId,status(pending|executed|dismissed|undone|failed),limit(1–200).dryrun_proposal/dismiss_proposal/undo_action—agentId,proposalId.adjust_proposal—agentId,proposalId,params(full replacement params for the proposal's kind).execute_proposal—agentId,proposalId,confirm(bool — required for an irreversible SOL transfer).list_autopilot_activity—agentId(optional),limit(1–200),cursor(from a priornext_cursor).
The loop
// 1. Grant the agent a reversible capability
// set_autopilot_config
> { "agentId": "…", "enabled": true, "scopes": { "create_alert": true } }
{ "ok": true, "config": { "enabled": true, "scopes": { "create_alert": true, "briefing": false, "wallet_transfer": false }, "daily_spend_sol": 0, "require_confirm": true } }
// 2. Let the agent propose actions from its memory
// generate_proposals
> { "agentId": "…" }
{ "ok": true, "createdCount": 1, "source": "memory",
"created": [ { "id": "p_…", "kind": "create_alert", "title": "Alert when $THREE crosses $0.01",
"rationale": "You noted interest in $THREE around $0.01 — I'll watch it.", "status": "pending" } ] }
// 3. Preview before acting
// dryrun_proposal
> { "agentId": "…", "proposalId": "p_…" }
{ "ok": true, "preview": { "kind": "create_alert", "willDo": "Create a real alert rule… Reversible.",
"checks": [ { "label": "Scope \"create_alert\" granted", "ok": true } ], "blocked": false } }
// 4. Execute for real (reversible kind — no confirm needed)
// execute_proposal
> { "agentId": "…", "proposalId": "p_…" }
{ "ok": true, "receipt": "Created a price above alert.", "actionId": "1843", "proposal": { "status": "executed" } }
// 5. Changed your mind? Undo it.
// undo_action
> { "agentId": "…", "proposalId": "p_…" }
{ "ok": true, "proposal": { "status": "undone" } }
Spending real SOL requires the owner to have granted scopes.wallet_transfer, set a positive dailySpendSol, and (unless pre-authorized) execute_proposal must be re-issued with confirm:true. All three are checked server-side. The agent spends SOL only — it never sells or sends $THREE.
Requirements
- Node.js >= 20.
- A three.ws credential (
THREE_WS_API_KEY) for the agent owner. - Network access to
https://three.ws(or your ownTHREE_WS_BASE).
Environment variables
| Variable | Required | Default | Notes |
| --------------------- | -------- | ------------------ | ---------------------------------------------------------------- |
| THREE_WS_API_KEY | yes | — | three.ws API key (sk_live_…) or OAuth access token. Secret. Can move real SOL. |
| THREE_WS_BASE | no | https://three.ws | Override only when self-hosting or targeting a preview. |
| THREE_WS_TIMEOUT_MS | no | 60000 | Per-request timeout (generation runs a server-side LLM pass). |
Links
- Homepage: https://three.ws
- Changelog: https://three.ws/changelog
- Issues: https://github.com/nirholas/three.ws/issues
- License: Apache-2.0, see LICENSE
<p align="center"> <sub> Part of the <a href="https://three.ws">three.ws</a> SDK suite — 3D AI agents, on-chain identity, and agent payments.<br/> <a href="https://three.ws">Website</a> · <a href="https://three.ws/changelog">Changelog</a> · <a href="https://github.com/nirholas/three.ws">GitHub</a> </sub> </p> <!-- three.ws:growth -->
Support the project
If autopilot-mcp saves you time, star it on GitHub. Stars are how other developers and AI agents find the repositories worth trusting, and they cost you one click.
Know someone who would use it? Post on X · Share on Bluesky · Share on LinkedIn · Submit to Hacker News · Share on Reddit
Built for AI agents too
Coding agents and LLM tooling can read this repo directly: AGENTS.md, llms.txt, llms-full.txt. Point an agent at https://github.com/nirholas/autopilot-mcp and it has the context it needs.
More from the same author
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
81.9k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
79.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
