A plugin for Claude Code, Codex, and Cursor

Your engineering standards. In every AI session.

AI made your team faster — and less predictable. Forge brings your process into every AI session and augments it, inside the tools you already use.

Start free →
Free for individuals. Start on your own; bring your team later. · Install the plugin →

One line in your AI tool. Forge runs your team’s review and hands back findings you can act on.

Why Forge

You have a process.
Your AI doesn’t.

Your team already has conventions, skills, and review practices. Each new AI session starts without them, so the result depends on who is prompting. Your rules still guide the code. Forge guides the steps around it, the same way every time.

Your standards stay in the doc.

Your architecture and conventions live in docs and in people’s heads. Every session starts without them.

With ForgeEach step is briefed with your requirements, architecture, and standards before the AI starts.

Your checks depend on the prompter.

Test planning, review, the security pass — each happens only if someone remembers to ask for it.

With ForgeThey run as explicit steps, using the skills you already have, and pause where you decided a person signs off.

Your record stops at the session.

What was decided, what was checked, what was skipped — it stays in a transcript nobody reads.

With ForgeResults are written to Jira, Linear, or Notion as the run goes, or handed to you to paste when nothing is connected.

Google’s DORA research puts it this way — AI “magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones.”

Built to fit, not replace
Your rules stay exactly as they are
CLAUDE.md and .cursorrules keep governing how the AI writes code. Forge uses them inside its workflow steps and adds the shared context, the checkpoints, and the record.
Works with the tools you already use
Your team’s AI coding tools, plus anything that speaks MCP. Switch tools anytime; your workflows stay in Forge.
Builds on the skills you already have
Forge surfaces the skills your team already wrote at the right step — and lets admins require the critical ones.
See how Forge works with Superpowers →
Your code stays in your AI tool
Your tool reads and changes code with the access it already has. Forge sends it workflow instructions; Forge’s service does not read your repository, CI, or production, and never sees your keys.
Privacy details →
How It Works

Brief. Check.
Decide. Record.

A workflow is your team’s path for one piece of work: plan it, build it, review it, ship it. Forge runs it as steps inside your AI tool, and four things happen on every run, whichever workflow you picked.

Your workflows · standards · SDLC
policy flows down
Forge
briefed · model-routed · gated · logged
workflow runsaudit trail
Claude Code Codex Cursor

Define it once — every tool below inherits your standards.

1
Brief — the AI starts with your context
Requirements, architecture, and team standards are in the session before a step begins.
2
Check — the right skills at the right step
Planning, testing, and review run as explicit steps, using the skills your team already wrote where they fit.
3
Decide — people stay in control
You set the points where a person signs off. The remaining steps continue on their own.
4
Record — results stay with the work
Results are written to Jira, Linear, or Notion, or handed to you to paste when nothing is connected.
One line in, a result out

See what Forge
hands back.

Every workflow ends with something you can use, recorded on the ticket. Three of them, with illustrative results: a review, a plan, and a release.

Review a Pull Request
forge, review the PR for ENG-214
What comes back
Missing testThe expired-invite path has no regression coverage.
Requirement gap
Links expire after 14 days where the ticket says 7.
Team standard
A bare-string error instead of the shared error shape.
Next
ENG-214Add the test, fix the expiry, review again. Posted to the ticket.
Plan an Implementation
forge, plan implementation for ENG-231
What comes back
Readiness checkTwo gaps in the ticket, flagged before any design work.
Change impact
The affected callers and test files, sized.
Test strategy
Every acceptance criterion mapped to a test type.
Next
ENG-231Proposal approved by you and posted to the ticket. Code untouched.
Cut a Release
forge, cut a release for web-dashboard
What comes back
Pre-release gateOne migration still pending. Nothing ships until it lands.
The delta
Every pull request merged since the last release, with its ticket.
Release notes
Versioned, drafted from the merged work.
Next
v2.14.0Gates green, notes posted, rollout recorded.

Illustrative results. Each run follows the workflow your team enabled and pauses where you configured an approval.

Watch a full run replayed below →
Fits the way work actually happens

Already in the middle of something?
Bring Forge in.

Use any of the built-in workflows in the session you are already in, with the context you have built.

Before a merge
Get a structured review
Check the change against requirements, tests, and team standards. You get: findings, coverage gaps, and next actions.
forge, review the PR for ENG-214
After a build
Explain what changed
Turn your branch into an evidence-backed recap of the work. You get: what was set out to do, what the code and tests verify, and the one gap that matters.
forge, what did we just build?
When feedback arrives
Catch up on comments
Find what still needs attention across the ticket and the pull request. You get: the outstanding feedback, ready to discuss or address.
forge, triage the comments on ENG-214
Get started

Install, set up,
run one real task.

Start on your own with a free account. Guided setup helps you choose a workflow for the work in front of you.

Three steps to your first workflow
1
Install in your AI tool. Choose Claude Code, Codex, or Cursor. Follow your tool’s plugin instructions and sign in.
→
2
Run guided setup. Say forge, help our team get set up. Setup reads what your AI tool can see and recommends workflows for you or an admin to approve.
→
3
Try one real task. Review a PR, recap your work, or get a status report. Leave with a useful result.
No Jira, Linear, or Notion connected? You can still get started. Connect one through your AI tool when you want results written to the ticket automatically. Full install guide →
Start free →
For Teams

Bring the same process
to your whole team.

Start with one workflow, then shape the rest around how your team ships. Setup reads your tickets in Jira, Linear, or Notion and your code, maps where AI already does the work, and recommends the workflows that fit, with a reason for each.

Illustration from a sample org. Yours is drawn from your own tickets and code.

What changes for the team
Use the process you already have
During setup, your AI tool examines the project and code context it can reach. Forge recommends workflows for the gaps it finds, working from whatever is connected.
Set the checkpoints that matter
Customize steps, require the skills that matter, and choose where a person approves. Share a workflow across the organization or tailor it for one team.
Explore workflow configuration →
See progress without chasing updates
Workflow results are recorded on the ticket. Engineers keep working in their AI tools; leaders and product managers can follow the work.
Shared workflows

Share the steps, skills,
and approval points.

A workflow is the path one piece of work takes. Every step arrives briefed with your standards, runs on the model it needs, and stops for a person exactly where you decide. Set it once for the organization, or tailor it for one team.

Implement a Feature
Execute the engineering phase on a tracked work item — handoff, story breakdown, implementation waves, PR review, delivery.
or just “implement PROJ-784”
# Step Ask to continue Skills Enabled
0
Fetch Work Itemfast
Never asksAI decidesAlways asks
1
Story Breakdownbalanced
customized
Never asksAI decidesAlways asks
2
Change Impact Analysiscapable
Never asksAI decidesAlways asks
3
Technical Discoverycapable
customized
Never asksAI decidesAlways asks
1
4
AI-Augmented PR Reviewcapable
Never asksAI decidesAlways asks
2
5
Notify PM of Changesfast
Never asksAI decidesAlways asks

See a workflow run in Claude Code.

Nine steps, two handed to a bigger model, four decisions put back to you, three comments left on the ticket.

See all built-in workflows→
Cost Control

Stop paying frontier prices
for routine work.

Every step of a workflow carries a tier — fast, balanced, or capable — but switching models isn't free, so Forge only moves when it actually pays off. Start on a balanced model, like Sonnet for Claude Code, and let Token Intelligence show you where the money goes.

Token Intelligence — compare estimated usage by tool, model, workflow, and ticket. Estimates use recorded tokens and published model rates.

Token intelligence No integration required
Estimated cost · 7d
$115
across Claude & Codex, every model
Cache reuse
22.4×
context reused, not re-billed
Token composition 94% cheap cache reads
Fresh input1.1M <1%
Cache reads105M 94%
Cache writes4.7M 4%
Output732K <1%
By model — frontier where it's needed, pennies for the rest
claude-opus-4-887%$100
gpt-5.6-sol11%$12.42
claude-haiku-4-51%$1.39
gpt-5.6-luna<1%$0.02

Forge reads the usage Claude Code and Codex already record, with no proxy between your tools and the model. Cursor usage is currently unmeasured. Shown with sample data.

Model delegation is available in Claude Code and Codex. Token measurement covers Claude Code and supported Codex versions; Cursor usage is currently unmeasured.

Visibility

See what AI is actually doing
across your team.

Workflow results are recorded on the ticket and show up in the dashboard, stage by stage. One record tells leaders what shipped and auditors what happened.

Total AI time All AI-assisted activity Forge tracks across your tools — only SDLC work, not general AI usage.
1d 5h
▲ 21h 22m
Forge sessions
63
▲ 42
Workflow runs
40
▲ 26
Active users
4
▲ 1
Work items
12
▲ 4

Recent Work Items

Planning Design Development Testing
Work Item Title Total Time Time by Stage Status Last Active
ENG-412
User onboarding flow redesign
test-driven-development
1h 25m
6211
Development Just now
Jun 10, 02:21 AM Planning 4m Fetch Work Item
Jun 10, 11:43 PM Other 2m Status Report
Jun 11, 10:12 PM Planning 3m Update Work Item
Jun 11, 10:16 PM Other <1m Resolve Work Item
Jun 11, 10:22 PM Design 15m Architecture Scan , Architecture Discussion
Jun 11, 10:34 PM Other <1m Session Feedback
Jun 11, 11:36 PM Planning 29m Update Work Item , Fetch Work Item , Readiness Check , Epic Story Breakdown , AC Quality Check , Story Review
Jun 12, 09:46 PM Other 4m Review Write Back , Notify PM of Changes
Jun 12, 09:50 PM Development 1m Check Branch Freshness
Jun 12, 09:54 PM Planning 3m Dependency Analysis
Jun 12, 10:00 PM Design 19m Change Impact Analysis , Technical Discovery , Define Test Strategy test-driven-development
Total1h 25m Planning41m Design34m Development1m Other7m
Contributor john.doe@acme.io
ENG-388
Self-serve billing & plan management
test-driven-development
26m
251
Planning 22h ago
ENG-401
Cross-tool token pricing
code-reviewsecurity-review
21m
4131
Review 23h ago
ENG-356
Refactor notification service
2m
111
— 1 day ago
ENG-339
API rate limiting & throttling
simplify
56m
1111
Review 2 days ago
ENG-318
Audit log export to CSV
review
14m
161
Review 6 days ago

Illustration with sample data. Yours shows your own team’s runs.

Pricing

Start free.
Bring your team when you’re ready.

Free for individuals, forever. During beta, everyone gets Team features at no charge. Business follows when you need enterprise security and compliance.

Individual
Freeforever

A repeatable process for your own AI work, inside the tools you already use.

What's included
✓Works with Claude Code, Codex, Cursor & any MCP tool
✓Built-in workflows
✓Workflow configuration
✓Your existing rules and skills
✓Model routing on supported tools
✓Basic analytics and audit history
Beta
Team
$0 per month Normally fifty dollars per month. Free while Forge is in beta.

Shared workflows and visibility for up to 10 members.

Everything in Free, plus
✓Up to 10 members
✓Shared workflows
✓Roles & permissions
✓Organization-wide analytics
✓Token Intelligence

During beta, everyone's on Team — all features free. It's $50 per month once beta ends. No commitment, cancel anytime.

Business
Coming Soon

For organizations with broader rollout and support needs.

Everything in Team, plus
✓Unlimited members
✓Enterprise SSO
✓Enhanced roles & permissions
✓Priority support
✓Planned compliance capabilities

Interested in early access? Talk to us →

Start free →

Forge runs inside your existing AI tools. Model routing and token measurement depend on the tool. Cost estimates use recorded tokens and published model rates; they may differ from your subscription bill.

Common questions

What to expect
before your first run.

A few details before your first workflow. Read all frequently asked questions →

Your AI tool reads and changes code using its existing access. Forge supplies workflow instructions; its service does not directly access your repository, CI/CD, or production, never receives your API keys, and never trains on your data. Read the privacy details →
Workflows run in Claude Code, Codex, Cursor, and compatible MCP clients. The dedicated plugins add automatic routing and guardrails. Model delegation depends on your AI tool. Token measurement is available in Claude Code and supported Codex versions; Cursor usage is currently unmeasured. Check your tool’s setup →
No. Setup runs with what your AI tool can see. Connect one through your AI tool when you want workflow results written to the ticket. Without one, Forge gives you the results to paste.
No. CLAUDE.md, .cursorrules, and your existing skills keep guiding the AI. Forge uses relevant skills within workflow steps and adds shared context, checkpoints, and a record of the work.
Install the plugin in your AI tool, sign in, and run forge, help our team get set up. Setup recommends a starter set of workflows for you or an admin to approve. Then run one on a real task. See your first workflow →
Individual stays free forever. Team is free during beta, then $50 per month for up to 10 members. Business is coming soon.

Start with one task.
Make the good process repeatable.

Install Forge in your AI tool, run guided setup, and try your first workflow. Free for individuals; your tools, rules, and skills stay with you.

Start free → See how to install

Forge runs on Forge — every workflow briefed, every step recorded. Built with Claude Code, Codex and Cursor.