Boundary infrastructure for the agentic enterprise.
Formael is the governed execution layer between AI agents and the external systems they act in. One protocol. One audit trail. One place to decide what an autonomous system is allowed to do in the real world.
Why this exists
The moment AI stopped just talking.
For years, the interesting question was whether AI could understand a request. That question is largely answered. The interesting question now is what happens when AI is allowed to act on a request - to send the contract, to issue the refund, to provision the account, to update the deal.
The instant a model can reach into a production system, the problem changes shape. Reasoning becomes a side effect. A wrong sentence is embarrassing; a wrong action is operational. And every team building agents quickly discovers that the hard part is no longer the model - it is the boundary the model crosses to do anything that matters.
That boundary doesn’t exist as infrastructure today. It exists as glue: hand-written API clients, ad-hoc credential management, prompt-level guardrails, spreadsheets of approval rules, and the quiet hope that nothing ever goes wrong. Formael is the conviction that this layer deserves to be a first-class piece of infrastructure, the same way load balancers and API gateways became first-class in earlier eras of compute.
What we build
A single, opinionated outbound gateway.
Every agent action passes through one place. That place authenticates the agent, resolves what it’s actually trying to do, evaluates organizational policy against it, optionally pauses for a human, executes against the right external system using credentials the agent never sees, and records the entire chain as evidence.
The agent never holds keys. The agent never learns one provider’s API and gets stuck. The organization never has to wonder what happened or who approved it. Switching providers becomes a configuration change, not a code change. And a whole fleet of agents can onboard through the identity they already carry — they present a token from your own provider, Formael provisions them under policy, and they immediately see only the capabilities their scope allows. Adding governance becomes a conversation about policy, not a sprint of refactoring.
That is the entire product. Everything else - the connector catalog, the policy engine, the approval workflows, the analytics - is in service of that one idea.
What we believe
Four convictions the platform is built on.
01
Intent over implementation
An agent should reason about goals, not endpoints. The boundary translates intent into the right deterministic call against the right system.
02
Governance is a choke point, not a sprinkle
Policy that lives inside agent prompts is policy that can be talked out of. Real control needs a single, un-bypassable surface to evaluate every action.
03
Audit is a first-class output
Every action produces evidence - what was attempted, why it was approved, what was executed, what came back. Append-only by default.
04
Speed comes from constraint
A good gateway lets teams ship faster precisely because it removes choices: one credential model, one approval shape, one trace format.
Where we are
Early, by design.
Formael is in early access. We work directly with a small number of design partners who are already building real agentic features into real products and want a serious answer to governance, audit, and connectivity from day one.
That shape is intentional. Infrastructure earns trust by being used in anger, not by collecting logos. We’d rather ship the right primitives for a handful of teams that depend on us than the wrong abstractions for a market we don’t yet understand. Pricing, packaging, and roadmap are all shaped in public conversation with the people running it in production.
Where we’re headed
Governance that learns from itself.
Every action that crosses the boundary leaves a paired record behind: what the agent was trying to do, and what the platform decided. The next phase of Formael reads that picture and helps you govern better — without ever moving the decision off a human. There are two AIs in an agentic system: the one that acts on behalf of the business, and the one that helps the people who govern it. We keep them apart. AI proposes; the policy engine decides.
Today
Governed execution
Every agent action runs the cycle - authenticated, weighed across four axes on the contents of the request, executed with credentials the agent never sees, and recorded permanently. Built and in production.
Next
Governance autopilot
Formael reads its own history to draft policy, explain verdicts in plain language, and simulate a rule change before you commit it. It advises your compliance and platform team - every suggestion versioned, reversible, applied only by a human, and running on the model you bring.
Later
A boundary both ways
The same control plane extends to inbound events, and eventually to agents transacting across organizational boundaries - each side enforcing its own policy. Further out, and deliberately so.
Get involved early.
If you are building agents that take real actions in production systems - or thinking carefully about how you will - we’d like to talk. Design partners get direct access to the team and meaningful influence on what we build next.
Early Access
Build the AI agents
you actually want to ship.
Join the Formael early access program. Get the boundary your agent fleet needs - governed execution, an append-only audit trail, and a neutral control plane you own - before you have to build it yourself.
Your Agent
Semantic intent envelope
Formael
Govern · Execute · Trace
External APIs
DocuSign · Jira · Slack…