Autospec CLI
Go CLI for spec-driven development: specify, plan, tasks, implement with YAML artifacts for Claude Code and Codex.
- 148 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs the agent on code edits with automatic commits and writes permissions to the agent configuration.
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/ariel-frischer/autospec/main/install.sh | shThe first Installation command from the README installs a prebuilt binary for Linux and macOS.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs the agent on code edits with automatic commits and writes permissions to the agent configuration.
This is third-party code. Review the repository files before installing.
What it does
`autospec` orchestrates spec-driven development: the `specify`, `plan`, `tasks` and `implement` stages run in one command, for example `autospec run -a "Add user authentication with OAuth"`. Each stage creates machine-readable YAML (`spec.yaml`, `plan.yaml`, `tasks.yaml`), checks it for completeness and retries the step on failure. Implementation runs in fresh agent sessions per phase or per task so context does not balloon. `autospec init` configures commands and scripts for Claude Code, Codex CLI, OpenCode or Jcode, and `autospec st` shows progress by phase and task. The idea grew out of GitHub SpecKit.
Who it is for. Teams and solo developers who run features through spec, plan, tasks, implementation and want to automate it instead of manual slash commands.
Good fit when
- A new feature needs a specification and plan before the first line of code.
- A long implementation eats context and calls for separate sessions per phase.
- Work status is read from `tasks.yaml` rather than chat.
- One process runs in turn on Claude Code, Codex and OpenCode.
Not a fit when
- The task is small: one request to the agent works without a spec.
- Windows without WSL: binary builds exist for Linux and macOS.
- Artifacts are wanted in Markdown, as in GitHub SpecKit: `autospec` stores them as YAML.
Example request
Prepare a specification and plan for a caching layer with `autospec run -sp` and stop before implementation.Limitations
Binaries are built for Linux and macOS (Intel and Apple Silicon); Windows users go through WSL. Installation runs the `install.sh` script via `curl`, and it is worth reading before running. `autospec init` writes permissions to the agent's global configuration by default, and a `--project` flag exists for the project. The `implement` command runs the agent on real code edits and automatic commits, and the `-y` flag skips confirmations, so work is better kept on a separate branch: `specify` itself creates a branch like `spec/001-user-auth`.
How to disable. Remove the `autospec` binary and the `specs/` directories with artifacts if they are not needed. Delete the permissions written by `autospec init` from the agent configuration.
FAQ
How do I check dependencies?
`autospec doctor` checks that an agent and git are installed.
How do I see progress?
`autospec st` shows artifacts and progress by phase and task.
How do I pick an agent?
`autospec init --ai codex` configures Codex, and several agents are comma-separated: `--ai claude,codex,opencode`. For a single run the agent is set by a flag of the `run` command.
How do I keep context small?
There are three implementation modes: one session per phase by default, one session per task, and one session for the whole spec.
How is it different from GitHub SpecKit?
Artifacts are YAML instead of Markdown, the CLI orchestrates stages instead of agent slash commands, and validation is programmatic with step retries.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI for spec-driven development: the agent carries a task through files in the repository from plan to verification and archive
FeatherSpec drives spec-driven development in Claude Code and GitHub Copilot: 13 /sdd commands and a Markdown project memory
Claude Code plugins for spec-driven development: specify, implement and review commands plus specialist agents
A terminal environment and CLI for spec-driven development: Gherkin, a task board and several agents, written in Rust.
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
A CLI, MCP server and skill for spec-driven development: short markdown specs, a task board and search that an agent can read.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Autospec (this card) | no | 148 | |
| Spec Kit | no | 149,633 | spec-kit installs commands and skills and the agent drives stages in Markdown; autospec orchestrates stages from the CLI and stores YAML |
| OpenSpec | no | 74,784 | openspec runs proposal, requirements and tasks as a process for the agent; autospec runs them automatically with validation and retry |
README in short
The project supports Claude Code, Codex CLI, OpenCode and Jcode, plus custom agent commands. A model can be assigned per stage, and settings sets switch via profiles.