HomeDocsWelcome

Start building with Agentic Fabriq.

Give your agents governed access to your tools — one MCP endpoint, per-user credentials, permissions enforced on every call, and a full audit trail.

Fabriq DeveloperYou are shipping a product with its own usersYour app holds an Agentic Fabriq app id and secret, and every call names one of your own end users. They never get a Fabriq account — you send them a connect URL, they authorise their own Gmail or Slack, and your agent calls those tools on their behalf.Who connects the tools: each of your end users, inside your product.Developer QuickstartDeveloper Agent PromptConnecting Your Users' ToolsTriggers & WebhooksStart with the Developer QuickstartFabriq EnterpriseYou are running agents across your organizationAgents act for the people in your own company. Members and teams sign in through your identity provider, admins grant per-user permissions on shared agents, and every action lands in one org-wide audit trail.Who connects the tools: your employees, in the Fabriq console.Build an Agent with AFGeneral Agent PromptSSO setupExternal-facing agentsStart with Build an Agent with AF
The BasicsSet up your first agent in minutes.MCP Client SetupPoint any MCP-compatible agent at Fabriq.SDK & CLIThe Python SDK and the afctl CLI.
Building with a coding agent?Drop the General Agent Prompt into Claude Code, Cursor, or CodexA copy-paste prompt that has your coding agent build the Fabriq integration for you — auth, tool listing, calls, and error handling included. Building on the developer SDK instead? Use the Developer Agent Prompt.

What you get

Everything your agents need, governed.

One endpoint, every tool.

Connect your agent to Fabriq once — every integration your organization has connected is available to it, behind one governed MCP endpoint.

your-agentclaude-code
⬡ Fabriq
GmailSlackGitHubPostgres

Auth per user, not per app.

Each person connects their own accounts. Credentials live in the vault and attach server-side at call time — agents never hold them.

MAYA@ACME · CONNECTIONS
Slack✓ connected
Gmail✓ connected
GitHub✓ connected
tokens live in the vault — injected per call

Permissions decided per call.

Default-deny grants resolve org → team → member, down to single actions — with holds for the sensitive ones. Changes apply on the next call.

read_tickets · allowedissue_refund $612 · needs_approvaledit_tickets · denied

Every action on the record.

Who asked, which agent acted, the tool, and the verdict — one audit trail across every agent, user, and tool.

14:02:11maya@ · ap-bot-01 · create_billallowed
14:03:05raj@ · support-bot · issue_refundneeds_approval
14:03:41dana@ · qa-bot-03 · export_docsdenied

Popular guides

How to Build an Agent with AFThe end-to-end guide, from registration to first tool call.General Agent PromptPaste into your coding agent — it builds the AF integration for you.Token Broker GuideThe alternative to MCP for agents that manage their own calls.Okta SSO SetupSign your users in through the identity provider you already run.External Users (B2B2C)Your product’s agents acting for your customers, per user.Custom Tools (API / MCP)Turn an OpenAPI spec or MCP server into agent-usable tools.Private Network ConnectorReach systems behind your firewall — outbound-only.Coding AgentsClaude Code, Cursor, and Codex on the governed endpoint.CLI Reference (afctl)Every afctl command, flag by flag.

Integrated tools

GmailDriveCalendarSlackGitHubNotionLinearStripeTwilioSalesforceZendeskAll connectors+ custom tools via OpenAPI & MCP

Need help?

Our team is here to help you get started.

Book a callContact us