01
Crews per task
Describe a task and Crewyard proposes who plans, implements, tests and reviews it, with the reason for each pick and a cost estimate. Edit the proposal in the Crew graph. Nothing starts until you approve.
Run crews of coding agentson your own accounts and keys.
Claude Code, Codex, Gemini CLI and OpenCode, supervised from one window. Crewyard proposes a crew for each task, watches limits and spend, and starts nothing until you approve.
01 What it does
One place for the agents you already pay for
02
Inbox
Agent questions, tool approvals and limit decisions in one place, with risk levels. High-risk commands are always approved one at a time, and an unanswered request keeps the run paused.
03
Limits and spend you can trust
Every quota or cost figure says whether it is Reported, Estimated or Unknown. Spend caps are per key and per task, and a run stops and asks at the cap.
04
Fallback and handoff
When a limit or a credit runs out, the task waits, asks, or hands over to a standby in the same worktree, by your rule for that task. Switching between two accounts of the same provider is opt-in.
05
GitHub under your control
Agents commit locally. A push queue, pull requests, merge gates and cleanup, and failing checks you can send to the agent after reading them. Every outward step waits for your click.
06
Task assistant and Autopilot
A read-only, terminal-style chat drafts tasks and research with you. Autopilot, off by default and per project, can run the task flow within limits you set; push, pull requests, merges and spend past a cap stay with you.
07
Your own agent files and roles
Share the instruction files, rules, skills and commands you wrote for Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Windsurf, Copilot or Cline with your profiles, as copies. Role files in .crewyard/roles change or add crew roles.
08
One bottom panel
The task assistant, your own terminal in the project folder or a task's worktree, the project's events, Crewyard's redacted logs, and every problem of the project with a button to where you fix it.
09
Reports and ready-made analyses
A finished research or review task shows its findings with severity, file and line, and suggests follow-up tasks as drafts. Ready-made analyses cover project review, optimizations, bugs, security, dependencies and licences, and test gaps.
02 How it works
From a task to a merged pull request, with you at each gate
- 01
Open a project
A folder, a clone or an empty repo. The wizard finds build and test commands and sets the rules.
- 02
Describe a task
Type it, attach images or files, or pick a ready-made analysis.
- 03
Approve the crew
Who does each step, why, with headroom and a cost estimate. You edit, then approve.
- 04
Agents work in worktrees
Each code-writing run gets its own git worktree; tests run at every turn boundary.
- 05
Answer in the Inbox
Questions, approvals and limit decisions wait for you, with risk levels.
- 06
Review, push and merge
Agents commit locally. Push, pull request and merge each wait for your click.
03 Inbox
Questions and approvals wait for you, with the risk spelled out
When an agent asks something, wants to run a command that is not allowed yet, or reaches a limit or a spend cap, its run pauses and the Inbox shows what it wants: the command, the working directory, the scope, the policy that applies, and who asked.
Approve once, allow it in this project, or deny with a reason the agent reads. High-risk commands are always approved one at a time, and a request never times out into an approval.
04 Runners and providers
Four agent CLIs, the models you choose
A runner is the agent program, a provider is where the model comes from, and a credential is a key or a plan sign-in; a profile combines one of each with a default model and its own config folder. Providers come from a synced catalog built from models.dev: Anthropic, OpenAI, Google, DeepSeek, Moonshot Kimi, Zhipu GLM, Alibaba Qwen, MiniMax, Xiaomi MiMo, OpenRouter and more. You can pin a catalog version so estimates stay on it, and compare its prices with LiteLLM's table.
Claude Code
Anthropic or Anthropic-compatible API keys only; never a claude.ai sign-in.
Checked with 2.1.312 or newer
Codex
ChatGPT sign-in through its own login, or an API key, on OpenAI or OpenAI-compatible providers.
Checked with 0.147.0 or newer
Gemini CLI
Driven over the Agent Client Protocol on a Gemini API key; its own Google sign-in serves Gemini Code Assist Standard or Enterprise accounts only.
Checked with 0.25.0 or newer
OpenCode
Runs on every provider in the catalog.
Checked with exactly 1.19.5, or exactly 2.2.4
05 Your call
What never happens without you
Agents work in their own worktrees and commit locally. Everything that leaves your machine or spends past a limit waits for an explicit click in the app.
- Starting a crew waits for your click.
- Pushing a branch waits for your click.
- Opening a pull request waits for your click.
- Merging waits for your click.
- Deleting a worktree or a branch waits for your click.
- Spending past a budget or a key's cap waits for your click.
- Continuing on another agent in Ask mode waits for your click.
- Answering an approval in the Inbox waits for your click.
Two opt-ins, both recorded
- Schedules you set up. A scheduled task goes through crew approval unless its template says to auto-approve that crew, and that choice is audited. Nothing runs while the app is closed.
- Autopilot. Off by default, per project, after a consent card. It approves crews only for tasks it created, under the per-task, per-day, hourly and concurrency limits you set. It has no tool for push, pull requests, merges, deletes, spend past a cap or Inbox approvals.
Get Crewyard
Built and tested on Windows 11. Every file, its SHA-256 and the install steps for each system are on the download page.