AI-powered API automation

The AI agent that
|

Give Forge your ERD, API spec and context. Describe a task. It plans and executes every API call — streaming results live.

Open Playground →See how it works ↓

5 free executions · no credit card needed

forge — plan + execute
Waiting to start...

Why Forge

Your product APIs are powerful.
Using them is the hard part.

Every product team has the same bottleneck: creating data across your own APIs takes a developer, a script, and half a day. Forge removes that bottleneck entirely.

✗The old way
✗Write seed scripts that break every schema change
✗2-day sprint to set up a QA environment
✗Customer Success opens engineer tickets for data setup
✗No visibility into what your own API actually does
✓With Forge
✓Describe the task in English — plan adapts to your current spec
✓5 minutes from ERD + Swagger to a full environment
✓CS types a brief, agent executes it — no ticket needed
✓Agent walks every endpoint and confirms it works end to end

How it works

Four steps from context to execution

📁
01
Upload context
ERD · API Spec · Context doc
›
💬
02
Describe task
Natural language instruction
›
🗺
03
AI plans
Decomposes into ordered API steps
›
⚡
04
Executes live
SSE stream · validates results

Real use cases

Built for teams that touch your APIs

Stop spending 2 days seeding demo data

A sales engineer at a B2B SaaS company needs a live demo environment by 2pm. The product database is empty. Normally that means filing a ticket, waiting for engineering, and praying the script still works after last week's schema migration.

With Forge: one task prompt → full product catalogue, customer accounts, and sample orders populated in minutes. No engineering ticket. No broken seed script.

6 min
with Forge
vs
2 days
manually
Example task prompt
"Set up a realistic demo store called
'Acme Demo' with 3 product categories,
12 products across those categories,
5 customer accounts, and 3 sample orders
in different fulfilment states
(pending, shipped, delivered).
Add discount code DEMO20 for 20% off."
✓ Forge generates a 24-step plan, gets approval, then executes all 24 API calls — streaming every result live.

Try it now

Ask Forge to plan anything

Plan Mode is a read-only Q&A mode — describe any task and Forge shows you
exactly how it would execute it. No side effects, no API calls to your product.

F
Hi! I'm Forge in Plan Mode. Describe any task and I'll show you the exact API steps I'd take — no side effects, no execution.

Plan Mode shows you the steps only — no API calls are made to your product. Open the Playground to execute plans →

Simulate

Stress-test your API with a crowd of AI agents

Spawn multiple agent personas simultaneously — power users, adversarial testers, read-only observers — and let Forge detect conflicts, race conditions, and broken assumptions automatically.

Builder agentsbuilder

Multiple builder personas run your product's typical workflows simultaneously: creating, reading, and updating resources just like real users would.

Adversarial agentsadversarial

Adversarial personas attempt deletes, overwrites, and edge cases. When a builder then fails to read a deleted resource, Forge surfaces the conflict immediately.

Observer agentsread-only

Read-only personas continuously poll your read endpoints. Any unexpected data change appears in the findings panel with a full event trail.

Try it in the Playground
Upload your ERD + Swagger, choose your persona mix (Auto or hand-pick), and run — no code needed. Results stream live with timestamped events and a downloadable conflict report.
Open Simulate tab →
Simulate
Agents
Scroll to see it run
Log

Start in 60 seconds

1
Upload context
2
Describe task
3
Watch it execute
Open Playground →Read the docs

Free tier · no credit card · 5 executions included

Got a question or use case?

I'm building Forge in public and genuinely want to hear from people using it — bugs, ideas, integrations, or just to say hi.