Getting started
Run your first DevSecOps review
Pick the DevSecOps template, describe your project, and walk through the live timeline to read your outputs.
Run your first procurement process
Use the Procurement Process team to go from requirements brief to contract heads of terms in under 30 minutes.
Set up your company profile
Why the company profile matters, what fields to fill in, and how it improves every future run automatically.
Ask your team for changes
Reply to a finished run and get a revised deliverable — the team keeps its context, and each turn becomes the next version.
Understanding run outputs
Navigate the timeline events, output hub, and confidence scores that appear on every completed run.
Find anything with Cmd+K search
Use global search to find runs by title or brief content from anywhere in the app.
Stay on top with the notification bell
The in-app bell shows live run updates. Never miss a completed run, blocked action, or checkpoint.
Organisation & billing
Manage your organisation
Invite team members, manage roles, and understand seat limits for your organisation.
Keep sensitive data out of outputs
Choose what happens when an output contains an email, a card number or an API key: flag it, redact it, or withhold it.
Upgrade your plan
How to subscribe, what each plan includes, and how to manage your billing via Stripe.
Understanding usage limits
How run and seat limits work, what happens when you hit them, and how to track usage.
Custom teams
Build a team with AI
Describe the outcome you want and let Forge assemble the roles, in order, as a starting team you then adjust.
Build a custom market research team
Step by step: pick agents, write the mission, set per-agent instructions, and run your first custom pipeline.
Get better output from a team
Three settings that change how a team handles its own weak spots: asking teammates, revising weak work, and bringing in a specialist.
Ground runs in your own documents
Upload your organisation’s material once, then have agents work from it on every run of a grounded team.
Let an agent use a site with no API
Give an agent a real browser for supplier portals, registers and internal tools — and fence it to the sites you intend.
Configure data tools for agents
Set up web search query templates, PDF reader, and HTTP fetch so agents can pull live data during a run.
Set per-agent instructions
What to write in per-agent instructions, worked examples, and how they override the default agent behaviour.
Quality, control & creators
Improve a team automatically
Use the Self-improvement panel to read a team’s quality trend, generate an optimised prompt, A/B test it, and apply the winner.
Steer a run while it’s running
Send a directive to a live run and watch it fold into the next agents — no cancel, no restart.
Publish teams as a creator
Claim a public handle, publish a team to the marketplace, and earn ratings, reviews, and a spot on the leaderboard.
Automation
Build a Workflow Chain
Chain several teams into an ordered pipeline that runs end-to-end, each step feeding the next.
Chain runs into a pipeline
Configure a Market Analysis to Investor Pitch chain so the output of run one automatically triggers run two.
Schedule recurring runs
Set up a cron-based schedule so Forge automatically fires a run on your chosen cadence — daily, weekly, or custom.
Trigger runs from GitHub
Set up an inbound webhook, configure the GitHub PR hook, and watch Forge start a DevSecOps review automatically.
Set up conditional Slack notifications
Use fire_condition syntax to post to Slack only when confidence meets your threshold, and hold lower scores for review.
Long-running and async runs
Pause a run for hours or days to wait for external data or a human decision, then resume automatically via callback, schedule, or snooze.
Integrations
Connect Jira to your agent team
Create Jira issues directly from agent findings. Configure the Jira tool, set field mappings, and test with a live run.
Use Forge from another AI assistant
Connect any MCP client to Forge so another assistant can run your agent teams as one of its own tools.
Use the public API
Generate an API key, create and list runs programmatically, and integrate Forge with your own systems.
Set up outbound webhooks
Register a webhook URL and receive a signed HTTP notification every time a run completes, fails, or is cancelled.
Verify webhook signatures
Validate the X-Forge-Signature header on incoming deliveries using HMAC-SHA256 to confirm authenticity.
Set up push notifications
Enable push notifications on mobile, install Forge as a PWA, and get notified the moment any run completes.
Build with the Developer Portal
Create a consumer app, configure scopes, deploy a team from a manifest, and make your first API call — all from the Developer Portal.
Deploy teams with Agent-as-Config
Define a full agent team in JSON and deploy it via POST /api/v1/teams so your application never needs the visual team builder.
Manage teams, schedules and automations as code
Keep teams, their schedules and the webhooks that fire them in one version-controlled forge.yaml, and reconcile with plan/apply.
Stream run events in the browser
Mint a short-lived stream token from your backend and use EventSource in the browser to show live agent progress without leaking your API key.
Use the Forge CLI
Install @forge/cli, authenticate, and create runs that stream live in your terminal — plus deploy teams and manage memory from the command line.
Run Forge in GitHub Actions
Add an automated AI review to every pull request: post the result as a comment and optionally gate the merge on the agent verdict.
Support
Get help with the support centre
Use the AI assistant for instant answers, open a ticket for human support, and track replies — all without leaving Forge.
Features
Cancel a run in progress
Stop a queued or running run before it finishes and understand what happens to partial outputs.
Save and reuse brief templates
Save your most-used briefs as named templates and launch them in one click.
Replay and compare runs
Re-run any completed run with the same brief and compare outputs side by side.
Annotate outputs with private notes
Add private notes to any agent output for review, follow-up, or handoffs.
Looking for technical reference?