wicked-studio
The ADE (agentic development environment) for running a team of coding agents: start a job with a chat, answer what needs you with the evidence, and build apps
What is wicked-studio?
The ADE (agentic development environment) for running a team of coding agents: start a job with a chat, answer what needs you with the evidence, and build apps and dashboards over your data in a chat.
Overview of wicked-studio
What it is
wicked-studio is an agentic development environment (ADE) for running a team of coding agents. It is a browser application that orchestrates Claude Code, Codex, GitHub Copilot, OpenCode, and Pi as a coordinated team. Users start a job with a chat; the system proposes a plan, and once confirmed, agents work in sandboxes while the user watches. With whisper.cpp installed, users can ask questions out loud and steer running jobs. Gate decisions are presented as a single card, and each job's evidence is a click away. The Learn feature suggests rules and memories from how the team works. The Apps feature turns databases, files, and APIs into dashboards built by chat. The wicked-crew daemon runs the jobs and bundles the studio.
Key features
- Job initiation via a catalog of workflows or a chat that proposes a plan; nothing starts until the user confirms.
- Agents are signed in once through the Agents page; jobs run each agent in a hardened sandbox (requires a container runtime such as Rancher Desktop, Docker, Podman, Colima, or OrbStack).
- Rules and memory are kept separate: Steering holds rules with explicit reach; Memory stores what happened and user instructions; agents suggest both, and the user approves.
- Human-in-the-loop gates surface decisions as a guided card with a recommended choice when available, a 10-second undo, and keyboard batch triage.
- Assurance receipts show what the run required, who built/evaluated/judged, what was skipped, and the attempt tree.
- Code graph building for onboarded repositories: blast-radius for any symbol, hotspots, a searchable call map, and a requirements view; Understand runs read-only analysis with cited answers.
- Live steering via the Desk: each run or chat opens as a session with a live record, gate cards, elicitation prompts, lifecycle controls, and step summaries.
Use cases
- Running governed coding agents on local or cloned repositories with optional PR delivery.
- Steering runs live by answering gate decisions, injecting messages, or retrying steps.
- Teaching the system rules and memories by approving agent suggestions or using the Learn feature.
- Building dashboards from databases, files, and APIs through chat.
- Understanding codebases through graph analysis and read-only question answering.
Getting started
See the repository README for installation and quick-start instructions.
License and pricing
Licensed under the MIT License. The project is open source and free to use.
Based on the project's public README and repository metadata.