How it works
Three steps. Two minutes.
Add to Slack
One click. Forge joins your workspace and works in any channel you invite it to.
Pick your AI & link a repo
Choose your AI provider — Anthropic, OpenAI, or OpenRouter — and paste your key in a secure modal. Then @Forge repo set owner/repo in any channel and Forge installs the GitHub App when needed.
Forge gets to work
Day-1 first-impressions brief on the repo you linked. Daily briefs, PR reviews, @mention answers, and memory all automatic from there. The longer it runs, the smarter it gets.
Features
Most AI tools forget. Forge compounds.
ChatGPT starts fresh every time. Forge builds memory every day.
Memory that compounds
Forge extracts decisions and key memories from every conversation. Semantic search finds the right one even when you phrase it differently. Nightly consolidation merges duplicates into richer insights. Month 1: helpful. Month 3: indispensable.
Morning briefs & weekly rollups
Forge posts proactive daily summaries and Monday rollups. It connects commits to conversations and flags what matters.
PR reviews in Slack
When a PR opens, Forge posts an instant review — what it does, what's risky, what to watch. With a clickable link to the PR.
Grounded in your code
Ask "what does this function do?" and Forge reads the actual file. It traces commits, references PRs, and cites code — not hallucinations.
100+ models, your choice
Bring your own key from Anthropic, OpenAI, or OpenRouter. Use Claude, GPT-4, Llama, Gemini, Mistral — whatever fits your team.
Workspace dashboard
See Forge's activity, memories, usage, and channel health at a glance. Search memories, manage your subscription.
Works in your editor
MCP support means Claude Code and Cursor can access Forge's project knowledge directly. No copy-paste from Slack.
Read-only access
Forge never pushes code, merges PRs, or modifies your repo. It reads and answers — that's it.
Pricing
Simple, transparent pricing
One plan. No tiers. No per-seat fees.
per workspace
Unlimited channels, repos, and users. You bring your own AI key — typical small-team usage is $3–15/mo on top, billed directly by Anthropic, OpenAI, or OpenRouter.
No credit card required
- Unlimited channels & repos
- 3 AI providers (Anthropic, OpenAI, OpenRouter)
- Choose your own model
- Workspace dashboard
- Full Slack + GitHub + MCP access
- Cancel anytime, no lock-in
Questions?
We've got answers.
Everything you need to know before getting started.
Still have questions? Reach out →Forge uses your own API key from Anthropic, OpenAI, or OpenRouter. You pay your AI provider directly — usually a few dollars/month for a small team. OpenRouter gives access to 100+ models including Llama, Gemini, and Mistral. The $49/mo subscription covers Forge itself, not the AI.
ChatGPT and Copilot are general-purpose tools that reset every session. They can't tell you why your team chose Redis over Memcached, who merged the auth refactor, or what broke in staging last Tuesday. Forge can — because it's been listening, reading your commits, and building memory since day one. Plus it's proactive: daily briefs, PR reviews, and stale PR nudges happen automatically. That accumulated context is something no chat-based AI can replicate.
Yes. Forge exposes project knowledge via MCP (Model Context Protocol), so Cursor and Claude Code can access team decisions and project status directly in your editor.
Messages in channels it's invited to, plus GitHub data (commits, PRs, issues, files) for repos you connect. Nothing else. Full privacy policy.
No. Forge is strictly read-only — it reads your code and conversations to give better answers, but never writes to repos or modifies anything.
Encrypted at rest with AES-256-GCM, used only for AI API calls on your behalf. You enter your key in a Slack modal — the value is never sent as a chat message and is never logged. Never shared with third parties.
Forge supports Anthropic Claude, OpenAI, and OpenRouter (100+ models) — you choose your provider during setup. Your messages and project data are sent to your provider's API using your own key. You provide and control your key, and you can switch providers or models anytime. DM Forge set model to change models on the fly.
Forge stores extracted key facts (memories) plus a rolling buffer of the last 100 messages per channel for short-term context. Full message history is not archived. Each workspace's data is completely isolated and never shared with other workspaces or used to improve Forge. After uninstall, your data is retained for 30 days then permanently deleted unless you reinstall. AI-generated responses may occasionally be inaccurate, so always verify critical information.
Yes. DM Forge set model to use any model your provider supports. OpenRouter users get access to 100+ models including Claude, GPT-4, Llama, Gemini, and Mistral.
Built by Strand & Stone — a small team working on cool stuff.