# Which agents work with Provely?

> Provely works with any agent through four surfaces: REST API, MCP server, CLI, and SDKs. One page per agent, with its guidance files and enforcement.

Canonical: https://provely.sh/agents  
Last reviewed: 2026-09-05  

**Any agent works with Provely. Every agent connects through four surfaces: the REST API, the MCP server, the CLI, and the SDKs. Each target is one profile plus guidance files.**

No agent forks the core. A new agent needs a profile and rendered guidance, not code.

| Agent | Kind | MCP | CLI | Hooks | CI gate | Verified on |
| --- | --- | --- | --- | --- | --- | --- |
| [Claude Code](/agents/claude-code) | coding agent | Supported | Supported | Supported | yes | 2026-09-05 |
| [Codex](/agents/codex) | coding agent | Verify at implementation | Supported | Verify at implementation | yes | 2026-09-05 |
| [Cursor](/agents/cursor) | coding agent | Verify at implementation | Supported | Verify at implementation | yes | 2026-09-05 |
| [Generic agent](/agents/generic) | agent | Supported | Supported | Not supported | yes | 2026-09-05 |
| [Grok](/agents/grok) | autonomous agent | Verify at implementation | Supported | Not supported | yes | 2026-09-05 |
| [Hermes](/agents/hermes) | autonomous agent | Supported | Supported | Not supported | yes | 2026-09-05 |
| [LangGraph](/agents/langgraph) | agent framework | Supported | Supported | Not supported | yes | 2026-09-05 |
| [OpenAI Agents SDK](/agents/openai-agents) | agent framework | Supported | Supported | Supported | yes | 2026-09-05 |
| [OpenClaw](/agents/openclaw) | autonomous agent | Verify at implementation | Supported | Not supported | no | 2026-09-05 |
| [pi](/agents/pi) | coding agent | Not supported | Supported | Verify at implementation | yes | 2026-09-05 |

## What does every agent do, on every surface?

1. `begin(contract, input)` returns an operation id and the correlation and idempotency metadata.
2. The agent does the side-effecting action with that metadata.
3. `action_result(operation_id)` submits the provider acknowledgement.
4. `verify(operation_id)` returns VERIFIED, PENDING, CONTRADICTED, FAILED, or UNVERIFIABLE.
5. `status` and `receipt` resume the operation later and fetch the signed receipt.

> The agent is untrusted. An agent report is E0, which is never sufficient. No surface gives an agent the verifier credentials.

## Read next

- [Verify Claude Code actions](https://provely.sh/agents/claude-code)
- [Verify Codex actions](https://provely.sh/agents/codex)
- [Verify Cursor actions](https://provely.sh/agents/cursor)
- [Verify Generic agent actions](https://provely.sh/agents/generic)
- [Verify Grok actions](https://provely.sh/agents/grok)
- [Verify Hermes actions](https://provely.sh/agents/hermes)
- [Verify LangGraph actions](https://provely.sh/agents/langgraph)
- [Verify OpenAI Agents SDK actions](https://provely.sh/agents/openai-agents)
- [Verify OpenClaw actions](https://provely.sh/agents/openclaw)
- [Verify pi actions](https://provely.sh/agents/pi)
- [Read the MCP server reference](https://provely.sh/docs/mcp)
- [See every verified service](https://provely.sh/verify)