Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Turn any repo into an agent-ready workspace for Claude Code, Codex, Cursor, and other coding agents.
| Date | Stars |
|---|---|
| 2026-07-24 | 1120 |
| 2026-07-25 | 1121 |
| 2026-07-28 | 1133 |
| 2026-07-30 | 1133 |
| 2026-07-31 | 1137 |
| 2026-08-06 | 1137 |
Today
— stars today
This week
+4 stars this week
This month
— stars this month
Momentum
39.0
growth rate 0.35%/day
# repository-harness Turn a software repository into a legible, agent-ready workspace. `repository-harness` gives coding agents a small entrypoint, structured repository knowledge, durable execution plans when work truly needs them, and mechanical validation. The repository—not a hidden workflow database—is the default system of record. The app is what users touch. The harness is what makes the app and its rules easy for agents and humans to understand. ## Why This Exists Coding agents commonly fail for ordinary engineering reasons: - important constraints live only in chat or in someone's head; - the repository does not say which documents are authoritative; - small changes are wrapped in process that obscures the actual work; - large changes lose decisions and progress between sessions; - validation is vague, late, or disconnected from user-visible behavior. The answer is not a longer mandatory workflow. It is a repository that exposes the right context at the right time and enforces important invariants with tests and scripts. This direction is anchored in OpenAI's [Harness engineering](https://openai.com/index/harness-engineering/) account: keep the agent entrypoint small, make repository knowledge navigable, store complex execution plans durably, make application behavior directly inspectable, and enforce architectural rules mechanically. ## The Default Workflow Start with [`AGENTS.md`](AGENTS.md), then follow the map in [`docs/WORKFLOW.md`](docs/WORKFLOW.md). The size of the request determines the amount of durable process: ```text read-only question -> inspect the smallest authoritative surface -> answer with evidence bounded change -> inspect locally -> change code or docs -> run relevant proof -> report the result multi-session or coordination-heavy change -> create docs/plans/active/<plan>.md -> record progress, decisions, and validation in Git -> move the finished plan to docs/plans/completed/ consequential ambiguity -> pause before mutation -> present the concrete choice and its effects -> continue after authority is clear ``` A typo fix does not need intake, a story row, or a trace. A migration spanning several sessions does need a durable plan. A request to “simplify permissions” without saying whether existing access may be revoked needs human judgment before code changes. These are independent decisions, not risk levels on one process ladder. ## Repository Knowledge - [`AGENTS.md`](AGENTS.md) — compact, stable entrypoint for agents. - [`docs/WORKFLOW.md`](docs/WORKFLOW.md) — canonical request and execution flow. - [`docs/HARNESS.md`](docs/HARNESS.md) — design principles and system model. - [`docs/ARCHITECTURE.md`](docs/ARCHITECTURE.md) — boundaries and dependency direction. - [`docs/product/`](docs/product/) — current product behavior and constraints. - [`docs/plans/`](docs/plans/) — active and completed durable execution plans. - [`docs/decisions/`](docs/decisions/) — durable architectural decisions. - [`docs/templates/exec-plan.md`](docs/templates/exec-plan.md) — plan template. - [`docs/README.md`](docs/README.md) — complete documentation map, including optional compatibility surfaces. - [`tests/README.md`](tests/README.md) — behavior ownership, validation entry points, and removal boundaries for test suites. The default path requires no local database. Product documents, code, tests, plans, decisions, and Git history form one inspectable source of truth. ## Install Harness Into A Project From a target project directory, run: ```bash curl -fsSL "https://raw.githubusercontent.com/hoangnb24/repository-harness/main/scripts/install-harness.sh?$(date +%s)" | bash -s -- --yes ``` On Windows PowerShell: ```powershell & ([scriptblock]::Create((irm "https://raw.githubusercontent.com/hoangnb24/repository-harness/main/scripts/install-harness.ps1"))) -Yes ``` Use `--merge` / `-Merge` to add missing Harness files without replacing existing project files. Use `--overrid
Excerpt of 9,645 characters
Read on GitHub
181
84
26
4
3
2
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:e1801bba5e59a694, topic:ai-agents
matched fp:e1801bba5e59a694, topic:context-engineering
matched fp:e1801bba5e59a694, topic:vibe-coding