What is Anvil?
Anvil is an AI developer tool and local-first, open-source workspace for running AI coding agents in parallel while keeping a human in control of what merges. You queue tasks against agents like Codex, Claude Code, and OpenCode, and each task runs in its own git worktree so nothing lands in your branch until you approve it. Quick fixes execute inline, while larger work tasks get a dedicated branch for review, which keeps parallel agent output isolated instead of colliding in one tree. Anvil adds a local memory system backed by Ollama embeddings or a self-hosted Postgres with pgvector to carry project context between tasks, and it lets you stack follow-ups so dependent work builds sequentially on earlier results. Real-time monitoring shows live output, tool calls, token counts, and cost per task, and a local analytics dashboard tracks tokens, costs, and task counts over 7-day, 30-day, and lifetime windows. The server runs on your host and is reachable from the desktop app, a LAN browser, or over Tailscale, with separate workspaces for different projects, credentials, and agent subscriptions. Licensed under Apache 2.0, Anvil is a self-hostable pick for developers managing multiple coding agents, listed in the BestAIFor directory.
What is Anvil used for?
Queuing and reviewing parallel AI coding agent tasks locally, with worktree isolation, project memory, and per-task cost and token tracking.
How much does Anvil cost?
Pros and cons
Pros
- Each task runs in its own git worktree and nothing merges until you approve it
- Queues parallel tasks across Codex, Claude Code, and OpenCode agents
- Local memory via Ollama embeddings or self-hosted Postgres with pgvector carries project context across tasks
- Real-time view of live output, tool calls, token counts, and cost per task
- Reachable from desktop app, LAN browser, or Tailscale, with isolated workspaces per project and credentials
Cons
- Local-first and self-hosted, so you run the server and manage your own agent subscriptions
- Value depends on already using agents like Codex, Claude Code, or OpenCode
- Local memory needs Ollama or a self-hosted Postgres with pgvector, adding setup
- No managed cloud or paid support tier is mentioned; it ships under Apache 2.0
Questions about Anvil
What is Anvil used for?
Anvil is a local-first workspace for running AI coding agents in parallel. Developers use it to queue tasks across agents like Codex, Claude Code, and OpenCode, keep each task isolated in its own worktree, and review changes before anything merges, while tracking tokens and cost per task.
Is Anvil free?
Yes. Anvil is free and open source under the Apache 2.0 license with no paid tier. You self-host the server, so your costs are the machine you run it on and the agent subscriptions you connect.
How does Anvil keep parallel agent work from conflicting?
Every task runs in its own git worktree, and nothing merges until you say so. Quick fixes run inline while larger work tasks get a dedicated branch for review, so multiple agents can run at once without colliding in a single tree.
Anvil vs running agents directly in the terminal?
Running an agent directly gives you one session at a time with no shared view. Anvil adds parallel task queuing, per-task worktree isolation, a local memory system, and live cost and token monitoring across a dashboard, so you manage several agents with human approval on merges.
What are the best alternatives to Anvil?
Reviews
Sign in to rate and review Anvil.
Sign in to reviewNo reviews yet. Be the first to review Anvil!