ForgeMind Runtime 2.0 — multi-agent execution engine

AI infrastructure for modern engineering teams.

ForgeMind orchestrates autonomous agents that plan, write, review, and ship code — across your repos, your CI, and your cloud.

forge › run/ship-feature
live
Run · LIN-2841

Add IP-based rate limiter

Initiated by Maya · 6 minutes ago · 5 agents

  1. Plan

    12s

    Decomposed into 6 subtasks · routed to 4 agents

    planner-01·opus-4.7
  2. Generate code

    1m 41s

    6 files · +184 −12 · type-checked

    codegen-03·sonnet-4.6
  3. Run tests

    18s

    247 specs · 184 / 247 passing

    tester-02·haiku-4.5
  4. Review

    Inline analysis & checks

    reviewer-02
  5. Deploy preview

    Canary on us-east-1

    deployer-01
codegen-03 is writingrate-limiter.ts
+38 −0
import { redis } from "@/lib/redis";

// token bucket — 100 req/min/IP
export async function rateLimit(req) {
  const key = `rl:${req.ip}`;
  const count = await redis.incr(key);

  if (count === 1) {
    await redis.pexpire(key, 60_000);
  }
  return { ok: count <= 100 };
}
@forge continue with deploy when tests pass…

Self-hosted available · SOC 2 Type II · Used by teams at Linear, Vercel, and Stripe

The platform engineering teams behind

Linear
Vercel
Datadog
Supabase
Stripe
Notion
Cloudflare
Sentry
Pipeline

From prompt to production, end to end.

One coordinated runtime moves work through every stage of your delivery pipeline — observable, gated, and reversible at each step.

  1. 01
    Prompt
  2. 02
    Plan
  3. 03
    Generate
  4. 04
    Test
  5. 05
    Review
  6. 06
    Deploy
  7. 07
    Monitor

Every step is a deterministic span. Every span is a row in your audit log. Every audit row is a place to gate, replay, or roll back.

01 — Build

Coding agents that work like senior engineers.

ForgeMind agents read your codebase the way your team does — they pick up your conventions, your style guide, and your import paths, and they ship code that looks like you wrote it.

  • Repo-native — no sandboxes
  • Reversible diffs with full provenance
  • TypeScript, Go, Python, Rust, Java
Read the docs
codegen-03·sonnet-4.6
writing 3 files
  • rate-limiter.ts+38
  • middleware.ts+41
  • rate-limiter.test.ts+62
import { redis } from "@/lib/redis";

// token bucket — 100 req/min/IP
export async function rateLimit(req) {
  const key   = `rl:${req.ip}`;
  const count = await redis.incr(key);
  if (count === 1) {
    await redis.pexpire(key, 60_000);
  }
  return { ok: count <= 100 };
}
02 — Review

Reviews that catch what reviewers miss.

Schema drift, perf regressions, ADR violations, security issues — surfaced inline with full context, before a human ever opens the PR.

  • Cites your ADRs and prior incidents
  • Plugs into required-status checks
  • Audit-ready trail per suggestion
Read the docs
#2841Add IP rate limiter
5 / 5 checks
apps/api/middleware.ts+4 −1
if (count > 100) return res.status(429).end();+const { ok } = await rateLimit(req);+if (!ok) return res.status(429).end();
reviewer-02commented on L47

Batch incr and pexpire in a Redis pipeline to halve round-trips.

Tests
247 / 247
Coverage
+1.4%
Security
0 issues
03 — Memory

A persistent memory of your engineering org.

Code, docs, ADRs, runbooks, and APIs become a typed semantic graph. Agents and engineers query it with the same primitives — and it stays in your VPC.

  • Re-indexed sub-minute on every commit
  • Typed graph, not bag-of-text
  • Private by default — no third-party egress
Read the docs
memory.search·"how does idempotency work in checkout?"
3 results · 142ms
checkout.tsADR-014webhooks.mdorders.tspayments.tsLIN-2841runbook
Top sources
  • ADR
    014 · Idempotency keys
    0.94
  • src
    checkout.ts
    0.89
  • doc
    Replaying webhooks
    0.81
The product

One control plane for every agent and every run.

ForgeMind operates like the rest of your stack — repos, runs, deploys, and observability under a single, calm surface.

Acme/platform-api/Runs
Runs

Active & recent

  • LIN-2841
    Add IP-based rate limiter
    Tests · 184 / 247
    Running62%
    6m 12s
  • LIN-2839
    JWK rotation for auth-service
    Reviewer-02 commenting
    In review
    14m
  • LIN-2837
    Idempotent checkout webhooks
    Merged to main
    Completed
    1h
  • INC-0142
    Bisect prod stack trace · checkout.ts:184
    Patch shipped — p99 142ms
    Completed
    3h
The platform

Five modules. One coordinated engineering OS.

Adopt any module standalone, or compose them end-to-end. Each is independently deployable, with the same API surface.

ForgeMind Build

Coding agents that ship production code.

Plan, write, refactor, and test code grounded in your repo's conventions. No sandboxes, no toy environments — Build runs against your real toolchain.

Read the build docs
codegen-03·writing rate-limiter.ts
+38 −0
import { redis } from "@/lib/redis";

// token bucket — 100 req/min/IP
export async function rateLimit(req) {
  const count = await redis.incr(`rl:${req.ip}`);
  if (count === 1)
    await redis.pexpire(key, 60_000);
  return { ok: count <= 100 };
}
Enterprise

Built for the security model your CISO already approved.

ForgeMind runs inside your perimeter. Agents inherit your identities, your scopes, and your audit trail — there's nothing new to negotiate.

01

Self-hostable

Runs in your VPC on AWS, GCP, or Azure — or fully air-gapped on-prem.

02

Audit-grade

Every prompt, diff, and deploy is signed and exported to your SIEM.

03

Identity-aware

SCIM, SSO, and per-repo RBAC. Agents act with the scopes you already issue.

04

Customer keys

Bring your own KMS. Encrypt artifacts, memory, and traces with your keys.

SOC 2 Type II
ISO 27001
HIPAA-ready
GDPR
View trust portal →
Customers

Engineering teams shipping real production software.

From banks to logistics platforms, teams use ForgeMind to standardize how their organizations write, review, deploy, and operate code.

ForgeMind replaced four point tools. The Runtime spans codegen → test → review → deploy in one observable trace — exactly what our SRE team wanted.
DO
Daniel Okafor
Director of Engineering · Everwell Logistics
The reviewer catches things our seniors miss — schema drift, perf budgets, and an honest read on whether a change matches our ADRs.
SL
Sofia Lindqvist
Principal Engineer · Halcyon Bank
We run ForgeMind in our own VPC against private GitLab. Same RBAC, same audit log, same KMS — it just slid into our security model.
HS
Helena Schroeder
Head of Infrastructure · Beacon Insurance
Pricing

Plans that scale from one engineer to one thousand.

Start free. Upgrade when your team is shipping with ForgeMind every day.

Developer

For solo engineers and side projects.

$0/ month
Start free
  • 1 connected repo
  • 200 agent runs / month
  • Build, Review, Memory
  • Audit log
    7 days
  • SSO / SCIM
  • Self-hosted runtime
Most popular
Team

For growing teams shipping daily.

$49/ seat / month
Start trial
  • Unlimited repos
  • 10,000 agent runs / month
  • All five modules
  • Audit log
    90 days
  • SSO (Google, Okta, GitHub)
  • Self-hosted runtime
Enterprise

For platform orgs with security needs.

Custom
Contact sales
  • Unlimited repos & runs
  • Dedicated runtime capacity
  • All modules + private models
  • Audit log
    Unlimited
  • SSO + SCIM + RBAC
  • Self-hosted / VPC / air-gapped

Bring autonomous engineering to your team this week.

Connect your repo, run your first workflow in under five minutes, and watch ForgeMind ship the first PR while you read the docs.