Codebuff
Terminal-based AI coding agent
Overview
Codebuff is a CLI coding agent — runs in your terminal, reads your codebase, makes multi-file edits, runs tests, iterates until things work. Free at low usage; subscription for higher limits. Strong reputation for not over-engineering small changes.
Use cases
What you can produce with Codebuff
- Describe a code change in plain English in your terminal and have the agent locate the relevant files, make multi-file edits, and apply them to your repo.
- Have the agent run your test suite or type checker after an edit and iterate automatically until the errors are resolved.
- Watch and click into individual sub-agents (planner, editor, reviewer) in the TUI to inspect what each one is doing mid-task.
- Switch between Free, Default, Max, and Plan modes to trade off cost against model quality for a given task.
- Define custom agent workflows and embed Codebuff agents in your own tooling via the open-source SDK.
- Install packages, scaffold new files, and execute shell commands as part of a single natural-language request.
- Use the ad-supported Freebuff tier to get daily free agent hours on budget models without a subscription.
ASEAN Perspective
Codebuff in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Codebuff is an open-source, terminal-first AI coding agent that understands your codebase and edits files directly from the command line, with an SDK for embedding it into projects. For developers who live in the terminal and want a transparent, scriptable alternative to IDE-bound assistants, it is a capable and refreshingly open option, with a free ad-supported tier (Freebuff) lowering the barrier to try it.
It suits CLI-comfortable developers and automation builders; those wanting a rich IDE UI will prefer Cursor or Copilot. Honest caveats: it is a smaller, younger project so the ecosystem and polish trail the leaders, credit/pay-as-you-go costs (around 1¢/credit) can add up on heavy use, and as an agent it needs review of its changes. The API/SDK and open codebase are genuine pluses. Usable worldwide from ASEAN in English.
What people say
Codebuff launched out of Y Combinator's Fall 2024 batch as a deliberately simple pitch: type what you want in the terminal, and the agent reads your codebase, edits files, and runs commands until the task is done. Since then it has evolved considerably — the product is now open source, built around an explicit multi-agent architecture that splits work between file-picker, planner, editor, and reviewer agents, with a TUI that supports mouse interaction and an SDK for embedding agents in your own workflows. In 2026 the company also launched Freebuff, an ad-supported free variant offering daily free hours on cheaper models.
Users who like Codebuff tend to praise exactly what the directory blurb suggests: it makes proportionate changes rather than over-engineering small requests, its whole-codebase file discovery is genuinely good, and the ability to run tests and iterate without leaving the terminal makes it feel like a real coding partner rather than a chat wrapper. The open-sourcing and the visible sub-agent activity have earned it goodwill among developers who distrust black-box agents.
The recurring complaints are about cost and marketing rather than capability. Pricing mixes signup credits, a 500-credit monthly free tier, pay-as-you-go at roughly one cent per credit, and the separate Freebuff offering — reviewers consistently describe this as confusing on first read, and long agent loops can burn credits unpredictably compared with flat-seat competitors. Its headline benchmark claims against Claude Code come from Codebuff's own eval suite, so skeptical reading is warranted. Setup is also heavier than some rivals since you still choose models.
Codebuff fits terminal-first developers who want an open-source, model-flexible coding agent and are willing to supervise its output and watch credit consumption. Teams needing predictable per-seat billing, or developers who want a zero-config experience, may find better-known alternatives like Claude Code or Aider a more comfortable default.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Codebuff's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Codebuff unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Codebuff directly →
Spotted something out of date? Suggest an update →
More in Code & Dev Tools
