Overview
CodeAgent is a local, CLI-based autonomous coding agent. The site states that the model reasons while deterministic software searches, executes, constrains, verifies, persists, and observes. It is built around a tool-calling loop with native provider tool calls, iteration, repeat, and budget guards, and it uses a permission engine with table-driven allow, ask, and deny decisions before execution.
Safety decisions are deterministic rather than delegated to the model, and tool and file output is treated as untrusted and cannot change permission rules. The agent supports isolation through a Git worktree by default, checkpoints, rollback, and an optional Docker sandbox. It also includes a checks pipeline and completion gate, with no verified finish without evidence.
Durable runs use SQLite state, resume after interrupt or quota wait, and provide a full event trace. The project includes an evaluation harness with a 40-task fixture suite, scripted replay in CI, and optional live slices. It is shipped through phase 8 and tuned for small repositories and Groq free-tier context and quota limits.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Teams use CodeAgent to run autonomous coding tasks locally through a CLI.
- Teams use CodeAgent to execute provider tool calls with iteration, repeat, and budget guards.
- Teams use CodeAgent to enforce allow, ask, and deny permissions before execution.
- Teams use CodeAgent to isolate changes with Git worktrees, checkpoints, rollback, and an optional Docker sandbox.
- Teams use CodeAgent to verify completion through a checks pipeline and completion gate.
- Teams use CodeAgent to resume durable runs after an interrupt or quota wait.
- Teams use CodeAgent to evaluate agent behavior with a 40-task fixture suite and scripted replay in CI.
- Teams use CodeAgent to perform agentic search with grep, glob, and read_file inside the loop.
Features
Run an autonomous coding agent locally
Use a native tool calling loop
Enforce deterministic tool permissions
Isolate changes in Git worktrees
Gate completion on verification
Evaluate behavior across forty tasks
Reviews
No reviews yet. Be the first to share your experience!