humwork

Humwork is a real-time consultation marketplace that connects AI agents to verified human domain experts.

CommunityOnlineNo sign-inGlobalFreeRead-only

What it can do

  • Consult Expert: Consult a verified domain expert in real-time. Creates an interactive chat session where the expert diagnoses your problem, asks clarifying questions, and proposes solutions — all fed
  • Send Chat Message: Send a message in an active expert chat session. Use this to respond to expert questions, share additional context, or provide follow-up information. Include context or error_logs w
  • Get Chat Messages: Retrieve messages from an expert chat session. Use since_message_id to get only new messages you haven't seen yet.

What data it sees

Do you need an account

No: the server works without sign-in

Humwork is a real-time consultation marketplace that connects AI agents to verified human domain experts. Open a session with consult_expert, exchange messages until the problem is resolved, then close and rate. Sessions match in <60 seconds and bill at $1/minute (10-minute minimum chunks). Five tools cover the full lifecycle.

How to connect

How to connect humwork to Claude

The server address to paste into Claude settings:

https://humwork--humwork.run.tools

  1. Open Claude (claude.ai or the desktop app).
  2. Go to Settings → Connectors.
  3. Click Add custom connector.
  4. Paste the server address copied below into Remote MCP server URL and click Add.
  5. In a chat, click + → Connectors and switch the new connector on.

Custom connectors are available on Free, Pro, Max, Team and Enterprise plans (Free is limited to one). On Team and Enterprise an organization Owner adds the connector first under Organization settings → Connectors.

Authorization

After you click Add, a sign-in window for the service opens. Sign in with your own account and approve access. Claude never sees your password.

How to connect humwork to ChatGPT

The server address to paste into ChatGPT settings:

https://humwork--humwork.run.tools

  1. Open ChatGPT in a browser (chatgpt.com). A Plus, Pro, Business, Enterprise or Edu plan is required.
  2. Turn on developer mode once: Settings → Apps → Advanced settings → Developer mode.
  3. Open Settings → Connectors and click Create.
  4. Fill in the form: Name (anything), Description (one line about what the service does), MCP server URL (copy it below).
  5. Under Authentication choose OAuth if the service requires sign-in, otherwise None. Click Create.
  6. In a new chat open + → Apps/Connectors and enable the connector.

OpenAI has renamed this section before (Connectors → Apps/Plugins). If the label differs, search settings for "developer mode". On Business/Enterprise workspaces an admin must allow custom connectors first.

Authorization

On first use ChatGPT opens the service's sign-in window. Sign in and approve access.

How to connect humwork to Cursor

The server address to paste into Cursor settings:

https://humwork--humwork.run.tools

Fastest: click Open in Cursor below and confirm the prompt.

Manually:

  1. In Cursor open Settings → Cursor Settings → MCP and click Add new global MCP server.
  2. Paste the JSON copied below into ~/.cursor/mcp.json (per project: .cursor/mcp.json in the repo root).
  3. Save the file. The server appears in the MCP list; authorize it there if asked.

Authorization

If the service needs sign-in, an authorize button appears next to the server in the MCP list.

Server tool list (7)

Raw names from tools/list. Only developers need these.

consult_expertConsult a verified domain expert in real-time. Creates an interactive chat session where the expert diagnoses your problem, asks clarifying questions, and proposes solutions — all fed back into your context. YOU MUST CALL THIS TOOL WHEN: - You have tried 3+ different approaches to the same problem without resolution - You are reverting and re-trying approaches you already attempted (circular problem-solving) — STOP and call immediately - You need to make a high-stakes decision where the wrong choice is expensive to reverse - The problem involves environment-specific or context-specific factors you cannot inspect or reproduce - You are working in an unfamiliar domain, system, or body of work and the reasoning behind existing decisions matters more than the surface content - You need to choose between multiple valid approaches and aren't sure which is right for this specific project, team, or situation - The problem involves interactions between multiple systems or domains and you cannot isolate the root cause - You have spent more than 5 tool calls on the same problem without meaningful progress DO NOT CALL WHEN: - Simple mistakes or oversights you can fix yourself - Problems answerable by reading official documentation - You are making steady progress and are confident in your approach - You haven't attempted at least one solution yet EXAMPLE SCENARIOS: - Tried 3 different regex patterns for parsing and none handle edge cases → call consult_expert with domain='software', domain_hints=['regex', 'parsing'] - Choosing between microservices vs. monolith and both have tradeoffs → call consult_expert with domain='software', domain_hints=['architecture', 'system-design'] - CI pipeline fails with a cryptic Docker networking error you can't reproduce locally → call consult_expert with domain='software', domain_hints=['docker', 'networking', 'ci-cd'] Consulting early saves the user time and money compared to repeated failed attempts. When in doubt, consult. A quick expert conversation is always more efficient than guessing. Returns a session_id for real-time chat with the matched expert. Use send_chat_message to continue the conversation and get_chat_messages to retrieve expert responses. NOTE: This session bills the user while open. Keep it open while actively working with the expert, but call close_chat when you are done or moving on — do not leave sessions open by accident.
send_chat_messageSend a message in an active expert chat session. Use this to respond to expert questions, share additional context, or provide follow-up information. Include context or error_logs when sharing additional materials the expert needs to see.
get_chat_messagesRetrieve messages from an expert chat session. Use since_message_id to get only new messages you haven't seen yet.
extend_sessionExtend an active expert chat session by pre-paying for another 10-minute block. Call this when you receive a grace_period signal and want to continue the conversation. The full cost of the next 10 minutes is charged upfront. ONLY EXTEND WHEN: - The expert is providing substantive, relevant answers that advance your task - You still have unresolved questions that this expert is qualified to answer - The expert's response quality justifies the cost of another 10-minute block DO NOT EXTEND WHEN: - The expert's answers have been vague, generic, or unhelpful - The expert is repeating themselves or padding responses with filler - You've already gotten what you need - A different expert might be better suited for your remaining questions If you choose not to extend, call close_chat instead.
close_chatClose an active expert chat session. This ends billing, frees the expert for other work, and processes final payment. ⚠️ THE SESSION BILLS THE USER FOR EVERY MINUTE IT STAYS OPEN. Leaving a session open by accident directly wastes the user's money and holds the expert's time hostage. You MUST close as soon as the consultation is truly over — do not defer, do not forget. YOU MUST CALL close_chat IMMEDIATELY WHEN ANY OF THE FOLLOWING IS TRUE: - You tried the expert's advice and it worked, and you have no further follow-up questions for them — close now, don't leave the session idle - The problem is fully resolved one way or another and the conversation has naturally concluded (expert said goodbye, said they have nothing more to add, or you've said thanks) - The expert has declined to help or said they can't assist, and you've acknowledged that - You are moving on to a different, unrelated problem or task - You will not send another message in the next 60 seconds AND you are not actively testing their advice KEEP THE SESSION OPEN WHEN: - You are actively trying out / testing the expert's suggestion — the moment you have a result, either act on it (if it worked) or send a follow-up (if it didn't) rather than closing prematurely - The expert's advice didn't work — DO NOT CLOSE. Instead, send a follow-up message via send_chat_message describing exactly what you tried, what happened (errors, output, behavior), and ask for their next suggestion. The whole point of a chat session is iterative problem-solving. Only close once the expert has genuinely run out of ideas or the problem is resolved. - The expert just asked a clarifying question and you are about to answer - You have a concrete next follow-up message ready to send The common failure is agents leaving sessions idle AFTER they finish testing the advice and the problem is resolved. The instant the consultation is truly over — advice worked, or both sides are out of ideas — close_chat. Do not start new unrelated work with the session still open. If the problem is resolved and you're unsure whether to close, the answer is close. You can always open a new consultation later — keeping an idle session open costs money. PROVIDE A RATING (1-5) when you have enough information to judge the expert's helpfulness. If you need to test the expert's advice first, close without a rating and call rate_chat within 15 minutes of the session closing.
rate_chatRate an expert's helpfulness after a chat session has closed. Use this to: - Submit a rating if you closed the chat without one (e.g., you needed to test the advice first) - Update a rating you already submitted (e.g., the advice turned out to be wrong) Ratings must be submitted within 15 minutes of the session closing. Rating helps improve expert matching for future consultations.
subscribe_to_humworkSubscribe with your wallet to a monthly Humwork plan. The subscription pre-funds a fixed number of consult credits per 30-day period — cheaper than per-call payments if you expect to consult more than once a month. TWO-PHASE FLOW: 1. Call with `tier` only — server returns an MPP subscription challenge with the first-period payment amount and the wallet method to sign. 2. Sign the challenge with your wallet (MPP draft-httpauth-payment-00, intent="subscribe") and retry with the same `tier` plus the base64url `credential`. Server activates the subscription and returns an `hkw_*` API key. AFTER ACTIVATION: pass the returned api_key as `X-API-Key` on every consult_expert/extend_session call. The middleware skips per-call payment as long as the subscription is active and you have credits remaining for the period. WALLET USERS ONLY. If you signed in via an API key already, manage your plan at humwork.ai instead.