Core
$495/mo
USD · billed annually
Think bigger.Ship it with AI.
Custom agents, private LLMs and automation — designed, built and run by one senior team.
9:41
Ops Reconciler Running · Forma Core 3.1
Reconcile last week’s invoices against POs and flag anything odd.
- Pulled 1,284 invoicesnetsuite.query
- Matched 1,262 to purchase ordersmatch.fuzzy
- Checked vendors against policypolicy.rules
- Flagged 22 anomalies for reviewrisk.score
Ask Forma anything…
+3,100 agents in production for 120 teams that measure what AI actually earns them.
- Northvale
- Kinetic
- Lumen Health
- Arcadia
- Brightfield
- Corvex
- Ostrava
- Pinehurst
Automate the busywork, amplify the people. Our systems turn messy operations into measurable growth and calmer teams.
We pair research-grade engineering with product thinking, so every model we ship has a job, an owner and a number it moves.
$38M
Revenue unlocked for clients through AI-led workflows.
18,200 agents running today
4×
faster from idea to production.
Median inference · 48ms
Real-time responses on dedicated, region-pinned clusters.
“Their support agent resolved 72% of tickets in month one — and our CSAT went up, not down.”
— COO, Lumen Health
// formaNew: Forma Core 3.1 ships region-pinned inference in Frankfurt, Singapore and Virginia — sub-50 ms for every enterprise plan.
Capabilities
We close the gap between promising demos and dependable production systems.
Intelligence, tailoredto how you work.
001
Agent architecture
Multi-step agents with tools, memory and guardrails — deployed on your infrastructure so data stays put.
- Tool use
- Evals
- On-prem
002
003
Theo Marchetti
Founder & Principal Engineer
Our vision
AI shouldn’t just take tasks off the list. It should give people back the hours and headspace to do their best work.
We combine engineering rigour with honest design, so the systems we build don’t just solve today’s problem — they open up the next opportunity.
Systems that scale with your ambition, built on the best model for each job — not the one we’re paid to resell.
- EN
- DE
- FR
- ES
- PT
- JA
- ZH
- HI
- AR
- KO
Semantic search
Hybrid vector + keyword retrieval for answers you can trace.
Unified context
One governed data layer feeding every model and agent.
Token-efficient
Caching and routing that cut inference cost by up to 70%.
95+ languages
Global deployments with locale-aware evaluation.
Teams who brought us in for one system and kept us for the roadmap.
Where human judgement and machine speed meet — a short film on how we work.
2 min watchIntelligenceby design.
// form 01 — Raw data
From raw data to refined intelligence, in four honest steps.
We are engineers, designers and researchers working at the edge of what AI can do — and the line of what it should.
Our story
Ships interpretable models that survive security review on the first pass.
Ines Duarte
Head of Machine Learning
Makes AI feel like a tool, not a black box — interfaces people trust.
Jonah Mbeki
Principal Design Director
Keeps p95 latency under a second across five continents.
Kenji Watanabe
Infrastructure Architect
Designs agent hand-offs around how your best people actually think.
Priya Nair
Lead Cognitive Scientist
Flexible tiers that scale with your business. No hidden costs — just systems that earn their keep.
Growth
$1,250/mo
USD · billed annually
Advanced agentic workflows.
- 10 automation flows
- Hosted vector DB
- 5 admin seats
- Priority support
Pro
Popular$2,900/mo
USD · billed annually
Custom neural architecture.
- Unlimited flows
- Custom fine-tuning
- 15 admin seats
- 24/7 shared channel
Scale
$7,500/mo
USD · billed annually
Common questions
Security, timelines, ownership and how we measure ROI — answered plainly.
Everything youneed to know.
We deploy into your cloud or on-prem, with SOC 2-aligned controls, region-pinned inference and no training on your data. Ever.
Field notes on frameworks, benchmarks and the decisions behind production AI.
All articles