# Documentation

Connect your agents, choose which actions need approval, and review their requests. Start with the overview to see how withHuman fits into your team's work.

Start here: [withHuman Overview](https://withhuman.ai/docs/getting-started/overview.md): The approval layer for AI agents.

## Agent skills

Install withHuman's agent skills so Claude Code, Codex, Cursor and other coding agents know how to set up and use withHuman:

```sh
npx skills add https://withhuman.ai
```

| Skill | What it covers |
| --- | --- |
| [`withhuman`](https://withhuman.ai/.well-known/agent-skills/withhuman/SKILL.md) | What withHuman is and how it works. withHuman is the approval layer for AI agents. It intercepts an agent's tool calls and decides, through rules and human reviewers, which run automatically, which are refused and which wait for a person to approve them. Explains agents, instances, approval requests and decisions, approval pipelines, escalation paths, the review queue, the hosted and self-hosted editions, and how to behave when your own tool call is held, denied or expired. Start here for any withHuman question; it points to the withhuman-cli and withhuman-pipelines skills for setup work. |
| [`withhuman-cli`](https://withhuman.ai/.well-known/agent-skills/withhuman-cli/SKILL.md) | Connect coding agents and other agent runtimes to withHuman with the withhuman CLI so their tool calls wait for human approval. Covers installing the CLI, guided onboarding, connecting a runtime to an existing agent, checking the connection, refreshing the tool catalogue, upgrading, and removing the adapter. Use for Claude Code, Codex, OpenClaw, Pi, Hermes, DeepSeek, Gemini CLI and Antigravity on macOS or Linux, hosted or self-hosted. |
| [`withhuman-pipelines`](https://withhuman.ai/.well-known/agent-skills/withhuman-pipelines/SKILL.md) | Decide which AI agent tool calls need human approval in withHuman, and who approves them, by writing approval pipelines and escalation paths through the withHuman MCP server. Covers connecting the MCP server, reading the current rules and the agents' tool catalogue, drafting branch, outcome, delegate and webhook blocks, conditions on tools and arguments, escalation paths with teams, levels and working hours, team escalation policies, webhook endpoints, previewing sample requests, activating and rolling back revisions. |

## Getting started

- [withHuman Overview](https://withhuman.ai/docs/getting-started/overview.md): The approval layer for AI agents.
- [Quickstart](https://withhuman.ai/docs/getting-started/quickstart.md): Connect a local agent, describe which calls need your approval, and decide your first request.
- [Set up MCP Gateway](https://withhuman.ai/docs/getting-started/set-up-mcp-gateway.md): Keep an MCP server's credentials in withHuman instead of on your agents' machines, and send every call through your approval pipeline.
- [Set up approval pipelines](https://withhuman.ai/docs/getting-started/set-up-approval-pipelines.md): Decide which of your agents' calls are approved automatically and which wait for a reviewer.
- [Connect agents built in code](https://withhuman.ai/docs/getting-started/connect-agents-built-in-code.md): Send the tool calls of an agent you build in code to withHuman for approval.
- [Connect Gemini CLI](https://withhuman.ai/docs/getting-started/connect-gemini-cli.md): Human in the loop approvals for Gemini CLI. Set rules for what runs automatically, and send the rest to your team to approve or deny.
- [Connect Antigravity](https://withhuman.ai/docs/getting-started/connect-antigravity.md): Human in the loop approvals for Antigravity. Set rules for what runs automatically, and send the rest to your team to approve or deny.

## Concepts

- [How withHuman works](https://withhuman.ai/docs/concepts/how-withhuman-works.md): The three core concepts: agents, approval pipelines and escalation paths.
- [Agents](https://withhuman.ai/docs/concepts/agents.md): What agents and instances are, and how they fit together.
- [Requests and decisions](https://withhuman.ai/docs/concepts/requests-and-decisions.md): How requests get a decision, and what happens next for the agent.
- [Approval pipelines](https://withhuman.ai/docs/concepts/approval-pipelines.md): How approval pipelines decide what happens next for a request.
- [Escalation paths](https://withhuman.ai/docs/concepts/escalation-paths.md): Who reviews a request, when they’re contacted, and what happens if nobody decides.
- [Immutability](https://withhuman.ai/docs/concepts/immutability.md): How resources keep their history, and how revisions let you make changes and roll them back.

## Using withHuman

- [Review queue](https://withhuman.ai/docs/web-app/review-queue.md): Review and decide approval requests that need a human.
- [Audit log](https://withhuman.ai/docs/web-app/audit-log.md): Explore the record of requests, decisions, and organization activity.
  - [Audit event types](https://withhuman.ai/docs/web-app/audit-event-types.md): Understand the events recorded in the audit log and the details they contain.
- [Agents](https://withhuman.ai/docs/web-app/agents.md): Connect agents and manage their instances and credentials.
  - [Provisioner tokens](https://withhuman.ai/docs/web-app/provisioner-tokens.md): Register agent instances automatically and manage provisioner tokens.
- [MCP Gateway](https://withhuman.ai/docs/web-app/mcp-gateway.md): Connect downstream MCP servers and let your approval pipeline decide every call through them.
  - [Gateway tunnels](https://withhuman.ai/docs/web-app/gateway-tunnels.md): Put MCP servers you run yourself behind approvals without exposing them or sharing their credentials.
- [Approval pipelines](https://withhuman.ai/docs/web-app/approval-pipelines.md): Define how requests are evaluated and when they reach human review.
  - [Conditions](https://withhuman.ai/docs/web-app/conditions.md): Choose which requests a branch, a Delegate block or an escalation path step applies to.
  - [AI Judge blocks](https://withhuman.ai/docs/web-app/ai-judge-blocks.md): Use an AI Judge to evaluate requests against a written policy.
  - [Webhook blocks](https://withhuman.ai/docs/web-app/webhook-blocks.md): Ask an external service to help decide an approval request.
  - [Webhook endpoints](https://withhuman.ai/docs/web-app/webhook-endpoints.md): Manage the endpoints that approval pipelines call for external decisions.
  - [Branch blocks](https://withhuman.ai/docs/web-app/branch-blocks.md): Run a set of blocks only for requests that match a condition.
  - [Delegate blocks](https://withhuman.ai/docs/web-app/delegate-blocks.md): Hand requests that match a condition to another pipeline.
- [Escalation paths](https://withhuman.ai/docs/web-app/escalation-paths.md): Choose who reviews requests and how they escalate over time.
  - [Teams](https://withhuman.ai/docs/web-app/teams.md): Choose who on a team is notified and how requests escalate.
- [Team administration](https://withhuman.ai/docs/web-app/team-administration.md): Manage teams, their members and the permissions they inherit.
- [Permissions](https://withhuman.ai/docs/web-app/permissions.md): Manage direct and inherited access, permission scopes and decisions outside routing.
  - [Permission reference](https://withhuman.ai/docs/web-app/permission-reference.md): All permissions, what they allow and where they apply.
- [Membership management](https://withhuman.ai/docs/web-app/membership-management.md): Invite people, manage member access and control who belongs to your organization.
- [API keys](https://withhuman.ai/docs/web-app/api-keys.md): Choose personal or organization API keys for scripts and integrations.
  - [Personal API keys](https://withhuman.ai/docs/web-app/personal-api-keys.md): Manage keys that let scripts act with your current permissions.
  - [Organization API keys](https://withhuman.ai/docs/web-app/organization-api-keys.md): Manage independent access and permission policies for shared integrations.
- [withHuman MCP server](https://withhuman.ai/docs/web-app/mcp-server.md): Let an AI agent set up approval pipelines, escalation paths, teams, agents and webhook endpoints for you.
- [Notifications](https://withhuman.ai/docs/web-app/notifications.md): Choose how withHuman reaches you for standard and urgent requests.
  - [Email](https://withhuman.ai/docs/web-app/email-notifications.md): Receive approval notifications at your account email or another confirmed address.
  - [Text message](https://withhuman.ai/docs/web-app/sms-notifications.md): Receive approval notifications by SMS at a confirmed mobile number.
  - [WhatsApp](https://withhuman.ai/docs/web-app/whatsapp-notifications.md): Receive approval notifications at your confirmed WhatsApp number.
  - [Slack](https://withhuman.ai/docs/web-app/slack-notifications.md): Receive personal approval notifications in your connected Slack workspace.
  - [Discord](https://withhuman.ai/docs/web-app/discord-notifications.md): Receive personal approval notifications through your connected Discord server.
  - [Phone app](https://withhuman.ai/docs/web-app/phone-app-notifications.md): Receive push notifications and review requests in the iPhone or Android app.
- [Integrations](https://withhuman.ai/docs/web-app/integrations.md): Manage your organization's connections to external services.
  - [Slack](https://withhuman.ai/docs/web-app/slack-integration.md): Connect your organization's Slack workspace and match members to their accounts.
  - [Discord](https://withhuman.ai/docs/web-app/discord-integration.md): Connect your organization's Discord server so members can receive direct messages.
- [Organization policies](https://withhuman.ai/docs/web-app/organization-policies.md): Manage organization-wide rules for reviews, request visibility and API keys.
- [Tool presentations](https://withhuman.ai/docs/web-app/tool-presentations.md): Make tool requests easier to read with clear titles and formatted arguments.
- [Account](https://withhuman.ai/docs/web-app/account.md): Manage your password and understand how you sign in to withHuman.
- [Billing](https://withhuman.ai/docs/web-app/billing.md): Manage your hosted plan, purchased seats, usage limits and payments.
- [Mac app](https://withhuman.ai/docs/web-app/desktop-app.md): Review requests and get notifications on your Mac.
- [Mobile apps](https://withhuman.ai/docs/web-app/mobile-apps.md): Sign in, receive notifications and review requests on iPhone or Android.
- [Sign-in and directory](https://withhuman.ai/docs/web-app/sign-in-and-directory.md): Connect your identity provider for single sign-on and manage how people join.
  - [Verified domains](https://withhuman.ai/docs/web-app/verified-domains.md): Verify company email domains and choose whether SSO admits new members automatically.
  - [Directory sync](https://withhuman.ai/docs/web-app/directory-sync.md): Keep members and teams up to date through your identity provider's directory.
- [Updates and telemetry](https://withhuman.ai/docs/web-app/updates-and-telemetry.md): Check for new releases and choose whether your deployment reports its version.

## API reference

- [API overview](https://withhuman.ai/docs/reference/overview.md): How the API is organized, how to authenticate, what an error looks like, and the rules every endpoint follows.
- [Agent Approval Protocol](https://withhuman.ai/docs/reference/agent-approval-protocol.md): What an adapter calls on behalf of a running agent. Create an instance and get its credential, ask for approval of a tool call, and wait for the answer.
- [Review queue](https://withhuman.ai/docs/reference/review-queue.md): The review queue is where approval requests wait for a person to approve or deny them. These endpoints list the queue, retrieve one request, claim it while you look at it, and record your decision.
- [Agent enrollment](https://withhuman.ai/docs/reference/agent-enrollment.md): Connect an agent from a developer machine. The agent asks for a code, a person approves it in the browser, and the agent exchanges the code for a credential. A withHuman extension, not part of the protocol.
- [Account](https://withhuman.ai/docs/reference/account.md): Who the session belongs to. Read the current member, their organization and permissions.
- [Agents](https://withhuman.ai/docs/reference/agents.md): Agents are the AI systems that ask for approval. Each agent has instances, and each instance holds a credential it uses to call the API.
- [Approval pipelines](https://withhuman.ai/docs/reference/approval-pipelines.md): The rules every approval request runs through, as an ordered list of blocks. A block can approve or deny a request, send it to a person, or pass it to the next block.
- [Escalation paths](https://withhuman.ai/docs/reference/escalation-paths.md): An escalation path decides who reviews a request once the pipeline hands it to a person, and what happens when nobody answers in time.
- [Webhook endpoints](https://withhuman.ai/docs/reference/webhook-endpoints.md): A webhook endpoint is a URL on a service of yours that a pipeline's webhook block can ask for an outcome.
- [Members](https://withhuman.ai/docs/reference/members.md): The people in the organization, their status, and their permissions.
- [Teams](https://withhuman.ai/docs/reference/teams.md): Groups of members. A team can be an escalation target, and its members inherit the team's permissions.
- [Invitations](https://withhuman.ai/docs/reference/invitations.md): Invite people to the organization by email.
- [API keys](https://withhuman.ai/docs/reference/api-keys.md): Personal API keys let a script act as you outside the browser: a whk_ bearer that carries a subset of your own permissions, changes with your permissions, and dies with your membership. Mint one, list yours, revoke it, and verify what a key can do.
- [Audit](https://withhuman.ai/docs/reference/audit.md): The organization's append-only log of what happened and who did it, and the request timeline it tells.
- [Permissions](https://withhuman.ai/docs/reference/permissions.md): Every permission a role can carry, grouped by area, with who can hold it and the endpoints that require it.
- [Schemas](https://withhuman.ai/docs/reference/schemas.md): Every named object the published API uses, with its fields. Type names throughout the reference link here.
- [OpenAPI schema](https://withhuman.ai/docs/reference/openapi.md): Read or download the OpenAPI contract for the published withHuman API.