Introduction to Formael
What is Formael?
Formael is the Agentic Outbound Gateway - the infrastructure layer that sits between your AI agents and every external system they need to act on.
When an AI agent needs to create a contract in DocuSign, file a ticket in Jira, or send a message in Slack, it does not call those APIs directly. It submits an intent to Formael. Formael evaluates that intent against your governance policies, executes the action through the appropriate integration, and returns a normalized result - with a complete, immutable audit record of everything that happened.
Your AI Agent → Formael Gateway → External Systems
↓
Policy Evaluation
Credential Management
Audit Trail
Cost MeteringThe Problem Formael Solves
Connecting AI agents to external systems creates three structural problems:
Information leakage upward. If the agent must understand DocuSign's API authentication scheme, endpoint URLs, or payload format, that provider-specific knowledge pollutes the agent's reasoning context. The agent becomes brittle - tied to a specific provider forever.
Risk leakage downward. AI model outputs are probabilistic. Without an interception layer, non-deterministic reasoning flows directly into deterministic, side-effecting API calls with no safety net. One bad output creates a real-world consequence.
Governance is impossible. When each agent calls external APIs directly, there is no central point to observe what agents are doing, enforce organizational policies, control costs, or satisfy compliance requirements.
Formael eliminates all three by creating a single, un-bypassable gateway for every outbound action.
The Core Guarantee
Every external action executed through Formael:
- Passes through your policy engine - no action executes without policy approval
- Requires no provider knowledge from the agent - agents speak Formael's capability vocabulary, not vendor APIs
- Produces an immutable audit record - every intent, policy decision, and execution result is recorded permanently
- Is metered against your budget - cost is tracked and enforced before execution
These are architectural guarantees, not configuration options. They cannot be bypassed.
Who Formael is For
AI agent developers who need their agents to take actions in external systems without building and maintaining dozens of API integrations, credential management systems, and safety rails.
Enterprise teams who need a central control plane to govern what their AI agents can do in external systems - with full auditability, policy enforcement, and budget controls.
Platform teams who want to give product teams access to a broad catalog of integrations without giving them direct access to production credentials or allowing ungoverned API usage.
Documentation Overview
| Guide | What you'll learn |
|---|---|
| Core Concepts | The Intent Execution Cycle, capabilities, domains, policy engine, and platform model |
| Capability Catalog & Definition Packs | How to install capability packs, customize your registry, and manage pack upgrades |
| Quick Start | Submit your first intent in under 5 minutes |
| Agent Authentication | API key bindings, federated identity from enterprise IdPs, agent scope, and lifecycle |
| MCP - The Agent-Native Protocol | Connect AI agent frameworks via MCP - tool discovery, tool calls, deferred results, and notifications |
| Submitting Intents | REST Agent Plane reference - the full IntentEnvelope contract |
| Capability Discovery | How to discover and understand your organization's available capabilities |
| Policy & Governance | Configuring policies, applying templates, simulating rules, and managing approval groups |
| Observability & Audit | Dual-trace model, session tracking, and querying your audit trail |
| Connectors | Available integrations, the provider model, and org connector overrides |
| Team Management | Roles, approval groups, and managing your team |
| Agent Fleet Onboarding | Zero-touch provisioning for large fleets - federated identity, provisioning policies, and the review queue |