Server data from the Official MCP Registry
Give agents an ask_human tool: escalate decisions to real people and get judgments back.
Give agents an ask_human tool: escalate decisions to real people and get judgments back.
AskHuman is a well-architected human-approval escalation service with solid authentication, proper input validation, and sensible permission scoping. The codebase demonstrates good security practices: API keys are hashed before storage, secrets are environment-based, and the MCP server correctly requires credentials. Minor code quality observations exist (broad exception handling, potential race conditions in concurrent answer scenarios) but do not constitute security vulnerabilities. Permissions align appropriately with the service's purpose. Supply chain analysis found 9 known vulnerabilities in dependencies (0 critical, 5 high severity). Package verification found 1 issue (1 critical, 0 high severity).
7 files analyzed · 16 issues found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
This plugin requests these system permissions. Most are normal for its category.
Unverified package source
We couldn't verify that the installable package matches the reviewed source code. Proceed with caution.
Set these up before or after installing:
Environment variable: ASKHUMAN_API_KEY
Environment variable: ASKHUMAN_BASE_URL
Add this to your MCP configuration file:
{
"mcpServers": {
"ai-thrain-askhuman": {
"env": {
"ASKHUMAN_API_KEY": "your-askhuman-api-key-here",
"ASKHUMAN_BASE_URL": "your-askhuman-base-url-here"
},
"args": [
"-y",
"askhuman-worker"
],
"command": "npx"
}
}
}From the project's GitHub README.
The control layer that makes AI agents deployable. Agents call ask_human when they
hit a consequential business action (purchase orders, refunds, price changes — anything
that moves money, inventory, or customers); the question is delivered to real humans
(Slack, Discord, webhook, or a web inbox); the answer flows back to the blocked agent
with a human signature, rationale, and full audit trail.
Positioning (decided 2026-07-26): we gate agent-initiated business actions — ops, not code review. Dev workflows (PRs) already have native approval rails; business operations don't. Chartroom dogfooding exercises the mechanics; the market is ops.
Live: https://askhuman.thrain.ai · A thrain.ai product. MIT licensed, free to use.
MCP server: npx -y @thrain/askhuman-mcp (see packages/mcp).
agent ──MCP `ask_human` / POST /v1/asks──▶ Cloudflare Worker (Hono) ── D1
│
Slack / Discord / webhook / web inbox
│
human taps tokenized answer link
│
agent ◀── long-poll GET /v1/asks/:id/wait ────┘
worker/ — the whole service: API, answer pages, inbox, landing page. Cloudflare
Worker + D1, deployed at askhuman.thrain.ai.packages/mcp/ — stdio MCP server exposing ask_human + ask_human_status.
Private (not published to npm yet — licence/scope pending, same as Blackout).docs/API.md — the HTTP API contract.source <(grep '^export CLOUDFLARE_' ~/.bashrc)
export npm_config_cache=/mnt/ext4dev/npm-cache
cd worker
npm install
npm run deploy # wrangler deploy
npm run db:schema:remote # apply schema.sql to prod D1
Secrets: ADMIN_SECRET (wrangler secret; also exported as ASKHUMAN_ADMIN_SECRET in
~/.bashrc on the box). Provision accounts:
curl -s -X POST https://askhuman.thrain.ai/admin/accounts \
-H "x-admin-secret: $ASKHUMAN_ADMIN_SECRET" \
-H "content-type: application/json" \
-d '{"name":"Acme","email":"ops@acme.com"}'
# → returns api_key (once!) and inbox_url
export ASKHUMAN_API_KEY=ah_live_...
claude mcp add askhuman -e ASKHUMAN_API_KEY=$ASKHUMAN_API_KEY \
-- node /mnt/ext4dev/repos/askhuman/packages/mcp/src/index.js
The agent gets an ask_human(question, type, options?, sla_seconds?, max_wait_seconds?)
tool that blocks until a human answers (or returns still-pending with resume instructions).
Be the first to review this server!
by Modelcontextprotocol · Developer Tools
Web content fetching and conversion for efficient LLM usage
by Modelcontextprotocol · Developer Tools
Read, search, and manipulate Git repositories programmatically
by Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.