// forward deployed engineering
Build it with us. Then it's yours.
An engineer from the team that built Functional AI builds or improves your agent with you, on the same evaluation, runtime and observability your engineers use, then hands over. Available on every plan, priced per engagement.
When our engineer leaves, nothing breaks.
the engagement
- 01
Workshop
Pick the first agent and write its contract: typed inputs, typed outputs, the tools it may call and the conditions on them.
- 02
Tools, dataset, Score Set
Register and test your endpoints, build the first dataset from real inputs, and generate the Score Set from the definition.
- 03
Build and iterate
Run experiments across models and repeats; fix what the findings point at until the trust score and the behaviour checks are where they should be.
- 04
Deploy
Deploy with your team holding the keys. Your app calls one endpoint; rollback is instant and the URL never changes.
- 05
Hand over
Every version, dataset and experiment is in your account. Your engineers own it from here, and can flag a bad call into a test in two clicks.
what you have at the end
- A versioned Agentic Function with a typed contract, deployed in your account
- Its datasets, Score Set and experiment results, so the next change is measured against this one
- Your tools registered with effects and guards, so the agent cannot act outside its remit
- A team that can change and deploy it without us
Prefer to build it yourself? Every plan includes the same platform; get started here.
Talk to an engineer
Tell us what you want to ship. We reply within 48 hours with a scope and a price.
// faq
Common questions
- What is forward deployed engineering at Functional AI?
- An engineer from the team that built the platform works with your team to get one agent to production, or to improve an agent you already run. They work in your Functional AI account on the same evaluation, runtime and observability your engineers use, so everything they build is yours to read, change and deploy.
- Does the engineer own the agent afterwards?
- No. The engagement ends with a hand-over: every version, dataset, Score Set and experiment is in your account, and your engineers can change and deploy the agent without us. When our engineer leaves, nothing breaks.
- Which plan do I need?
- Any plan. Forward deployed engineering is available on Pro, Growth, Scale and Business, priced per engagement, and Enterprise includes engineering hours.
- What does it cost?
- Engagements are fixed-scope and priced per engagement: getting one agent to production, or improving an existing one. Tell us what you want to ship and we reply within 48 hours with a scope and a price. Platform usage is billed on your plan as usual, and your model provider bills tokens directly.
- Can you improve an agent we already have?
- Yes. Improving an existing agent is the second engagement shape: we define its contract, build a dataset from your production traffic and flagged failures, write the Score Set, and iterate until the experiments show the new version is better than what is live.
- Do you run the agent for us?
- The platform runs it. The engineer builds and improves the agent and hands it over; production hosting, versions, rollback and tracing are part of Functional AI on every plan, not a service.