AnyForge Code Studio

Single-agent AI coding.
Governed by default.

A browser-based AI coding agent with 200+ models via OpenRouter, native GitHub integration, and AnyForge Control built in. Escalate to a governed Crew when the scope outgrows a single agent. Stay on the conversation. Approve HIL gates in-thread. Free for everyone — bring your own provider key, AnyForge charges $0.30 per 1M tokens on the token flow.

Try Code Studio free →Prefer to keep your own IDE? See Control →

Powered by AnyForge Control

Every LLM call Code Studio makes flows through AnyForge Control. You get per-call cost analytics, intent-based routing to the right-sized model, portable memory across providers, and a hash-chained audit log — built in, nothing to install separately.

Features

Built for working engineers. No local setup.

200+ models via OpenRouter

Claude, GPT, Gemini, Llama, Mistral, and more. Switch models per request. No lock-in, no re-configuration.

Native GitHub integration

Commit and push from the browser. Main branch stays protected. No local Git setup.

Repo-grounded plans

Plans reference your actual codebase. Surgical edits, not generic patterns. Built-in terminal runs builds and tests.

Quality-first testing

Writes failing tests before fixes. Regressions caught before they ship. Unprompted vulnerability scanning on every change.

Control is native

Per-call cost analytics, intent-based routing, portable memory, and hash-chained audit come built in. Nothing extra to install.

One-click escalation to Crew

When scope outgrows a single agent, Code Studio proposes a governed Crew inline. Stay on the conversation. Approve HIL gates in-thread.

Codebase intelligence

Reverse-engineer any repo into a full documentation suite — architecture diagrams, security threat model, tech debt registry. Gap findings auto-seed DRAFT initiatives.

Codebase Intelligence

Understand any codebase in minutes.
Ship with confidence from day one.

Connect a GitHub repository and Code Studio reverse-engineers it into a complete documentation suite — architecture diagrams, security gaps, tech debt, runbooks, and more. Every finding auto-seeds a specific, actionable DRAFT initiative you can launch into a governed Crew. Two commands. No manual diagramming. No week-long archaeology sessions.

Phase 1 · 2–5 min

Repository Reconnaissance

  • →Structural Map — architecture, entry points, module inventory
  • →Gaps & Assumptions — open questions as a checklist
  • →Well-Architected Scan — reliability, security, cost, ops
  • →Gap initiatives auto-seeded to Initiatives page

Phase 2 · 10–20 min

Full Documentation Suite

  • →C4 architecture diagrams — L1 / L2 / L3 interactive drill-down
  • →Security threat model + data model + API overview
  • →Deployment topology + operations runbook
  • →Onboarding guide + tech debt registry + executive summary

Architecture Browser — live at /codebases/[id]/architecture

Every generated document is browsable in a sidebar-navigated viewer. C4 diagrams are interactive — click any component to drill from system context down to container and component level. Share the URL with your team as a living reference that stays current each time you re-run analysis.

Pricing

BYOK. No markups on model costs.

Bring your own API keys. Code Studio is free to start — no subscription, no seats, no daily caps. Sign up for a 100M-token free trial allowance (typically 1–2 weeks of active dev use before any AnyForge bill); after that, a flat $0.30 per 1M tokens on what flows through the harness — provider- agnostic, the same rate whether you use Anthropic, OpenAI, Bedrock, or self-hosted Ollama / vLLM. LLM costs go straight to your provider at wholesale. Full cost visibility per call.

Self-serve

100Mfree tokens on signup

  • →No subscription · No seats · No daily caps
  • →~1–2 weeks of active dev use before any AnyForge bill
  • →First Crew on us — 20K-token trial, one per company
  • →Up to 16K output tokens per request, 35 tool rounds
  • →Every model — your BYOK key, your contract

Platform fee

$0.30per 1M tokens, any provider

  • →Same flat rate for cloud BYOK and self-hosted (Ollama / vLLM)
  • →Routing + caching typically save more than the platform fee costs
  • →Unified usage view across products
  • →Separate AnyForge fee + model spend line items

Enterprise

Customannual contract

  • →Same $0.30 / 1M, monthly post-paid
  • →SOC 2 attestation · SSO
  • →Audit-log retention · SLA
  • →Self-hosted / regulated deploy

Code Studio builds it. Crew governs it.

Start in a single-agent thread. Escalate to a governed multi-agent Crew when the data or the scope says to. Same conversation, same token ledger, same audit chain.