Humanloop for AI agents
Humanloop helps developers build and refine AI applications with feedback loops and experiment tracking on model prompts. Product and engineering teams use it as a building block inside their own AI features. Connect it once through Arc0, and your agent, or Claude, ChatGPT and Cursor, can use it through one MCP endpoint, limited to what each user approved.
What agents do in Humanloop.
List experiments on a project
List active experiments on a project to see which prompt variant is being tested.
Check session history
List sessions for a project to review how real users interacted with a prompt.
Delete a project with confirmation
Delete a project only after the owning team confirms it is no longer active.
4 Humanloop actions, graded by risk.
Every Humanloop action is tagged read, write or destructive, so one policy covers the whole app and new actions inherit the right default.
read
2Look things up. Allowed by default.
- humanloop.list_sessionsList Sessions
- humanloop.list_experimentsList Experiments
write
1Create and change things. Allow, or ask the user first.
- humanloop.create_projectCreate Project
destructive
1Delete, cancel or archive. Ask first, or deny outright.
- humanloop.delete_projectDelete Project
Humanloop in three steps.
- 01Your users connect HumanloopThey add their Humanloop api key on Arc0 Connect, under your brand. It goes straight into the vault.
- 02You set the rulesReads run, writes like “create Project” can wait for the user, and “delete Project” can be denied outright.
- 03Any agent can actYour agent calls Humanloop through the Arc0 SDK or MCP, and so can Claude, ChatGPT and Cursor. Every call lands on the audit log.
await arc0.policies.set('humanloop', { read: 'allow', write: 'ask', // create_project destructive: 'deny', // delete_project }) # Claude Code: the same connection, one URL $ claude mcp add --transport http arc0 \ https://mcp.arc0.ai/u/u_8f2
How Humanloop connects.
Users add their Humanloop api key on Arc0 Connect. It is encrypted in the vault, never shown to the model, and each user can rotate or revoke it at any time.
The same Humanloop connection serves your agent over MCP and your own backend over REST and the proxy, so a user connects once. How Arc0 handles credentials →
- AUTH
- API key
- CREDENTIALS
- Per-tenant encrypted vault
- MODEL SEES
- Results only, never credentials
- AUDIT LOG
- Every call, on every plan
Use Humanloop from any agent.
ClaudeChatGPTCursor
Codex
VS CodeOpenAI Agents SDKClaude Agent SDK
Vercel AI SDK
Mastra
LangGraph
Humanloop and Arc0, answered.
Can I use Humanloop with Claude, ChatGPT or Cursor?
Yes. Connect Humanloop to Arc0 once, then add your Arc0 MCP URL to Claude, ChatGPT, Cursor, Claude Code or any other remote-MCP client. Each assistant only gets the Humanloop actions you allow.
How do users connect Humanloop?
Users add their Humanloop api key on Arc0 Connect. It is encrypted in the vault, never shown to the model, and each user can rotate or revoke it at any time.
Which Humanloop actions can my agent take?
4 in total: 2 read, 1 write and 1 destructive, such as “create Project”. Your policies decide which of them each agent may call.
Can I stop my agent from deleting things in Humanloop?
Yes. Actions like “delete Project” are graded destructive. Set destructive actions to deny, or to ask so the user approves each one, and blocked calls still show up on the audit log.
Can my own backend call Humanloop too?
Yes. The same Humanloop connection is available over REST and through the Arc0 proxy, so your product and your agent share one connection per user.
Plug Humanloop into your agent.
Your users connect Humanloop once, under your brand. Your agent gets 4 actions behind your policies, with every call on the record.
Free to build · MCP + REST · Audit log on every plan