57 AI Agents · 10 Phases · Self-Evolving · Chat Channels

Tickets in.
Software out.

Anvil orchestrates 57 specialized AI agents to turn your Jira, GitHub, and Trello tickets — or your Slack and WhatsApp messages — into fully implemented, tested, reviewed, and deployed code.

Free tier: 200K tokens/month. No credit card required. See Pricing
59+
Specialized agents
10
Pipeline phases
<1s
Machine boot time
$0
Free plan, all agents

How It Works

Three steps from ticket to deployed code.

1

Connect Your Sources

Link your Jira, GitHub Issues, or Trello board. Connect Slack, WhatsApp, or Telegram for chat-driven development. Anvil listens via webhooks.

2

Agents Do The Work

57 specialized agents plan, code, test, review, and check compliance. They learn from your feedback and improve over time. All in ephemeral Fly.io machines.

3

PR Delivered

A tested, reviewed, compliant pull request appears in your repo. Optionally auto-deploy via Forge. Results sent back to your chat channel.

Built for Autonomous Development

Everything you need to automate your development pipeline.

57 Specialized Agents

From ResearchAgent to DeployAgent. Each agent has a specific role, system prompt, and structured output schema.

10-Phase Pipeline

Planning, diagnosis, implementation, quality, review, compliance, PR, DevOps, AI/ML, and communication.

Configurable Approval Gates

Supervised, semi-auto, or full-auto. You decide how much human oversight each project needs.

Sentry Auto-Fix

Proactive monitoring detects production errors. Anvil debugs, writes the fix, and opens a PR automatically.

OpenTelemetry & Datadog Ingestion

Pipe production stacktraces straight into the agent pipeline. Point your OTel exporter at Anvil, or wire up Datadog Error Tracking — no vendor lock-in, same auto-debug → PR flow as Sentry.

Agent Memory & Evolution

Agents remember context across interactions and improve their prompts based on your feedback. Dream Mode consolidates learnings overnight.

Chat Channels

Connect Slack, WhatsApp, Telegram, Discord, or Zoom. Chat with conversational agents — or escalate to the full pipeline.

11 Stacks Ready

Laravel is live in production. Django, Next.js, Rails, Spring Boot, Express, Angular, SvelteKit, Nuxt.js, Astro, and FastAPI ship with complete per-stack prompt overrides and Dockerfile scaffolds — same 59 agents, same pipeline, idiomatic per-stack guidance. Activation is publishing the image. No platform rewrite to add a language.

Real-Time Streaming

Watch agents work in real-time via WebSocket. See every step as it happens.

Conversational Agents

Team Assistant, Meeting Intel, Sprint Reporter, Code Explainer — pre-built templates for real-time team operations via chat.

The Agent Pipeline

Every ticket flows through 10 sequential phases.

8 agents
Planning
BLOCKING
1 agent
Diagnosis
17 agents
Implementation
9 agents
Quality
BLOCKING
3 agents
Review
3 agents
Compliance
BLOCKING
1 agent
PR Creation
5 agents
DevOps
2 agents
AI/ML
5 agents
Communication

Security First

Three layers of isolation. Zero trust architecture.

API Key Isolation

Your Anthropic API key never leaves Anvil's servers. Agent machines receive a proxy URL — no key, no credentials.

Ephemeral Compute

Every agent session runs in a fresh Firecracker micro-VM. Destroyed after completion. Zero persistent state.

Database Isolation

Every tenant gets their own database. Connection-level isolation. A bug can't leak data across tenants.

Ready to automate your development?

Start with the free tier. 200K tokens/month. All 57 agents. No credit card.