Charter

AI-agent readiness, scored.

8 followers

AI-agent readiness, scored.

8 followers

Charter is an offline-first CLI that audits any repo against 18 rules and returns a deterministic 0–100 score in under 2 seconds.

Free

Launch tags:Developer Tools•GitHub

Launch Team

Render Sandboxes: a computer for every agentIsolated compute, persistent state and a real filesystem

Promoted

Maker

📌

Hey Product Hunt — I’m Tashfiq, the maker of Charter. I built Charter after repeatedly seeing AI coding agents fail in repositories for reasons that were not really model failures. The repo had missing context, unclear conventions, unpinned tools, unsafe config, weak test paths, or secrets sitting somewhere an agent could easily read. The agent would still try to help — but the repo had not given it a safe environment to work in. That made me wonder: why do we have linters, test coverage, CI checks, and security scanners, but no simple number for “how ready is this repository for an AI agent?” Charter is my attempt at that number. It’s an offline CLI that scans a repo and gives it a deterministic 0–100 agent-readiness score across areas like context, secrets, MCP safety, environment, CI, testing, autonomy, and governance. No LLM in the core. No network calls. Same repo, same score, every time. The approach evolved from “maybe this should be an AI review” to the opposite: a boring, reproducible rule engine with clear fixes. I wanted something teams could trust enough to run locally, put in CI, and improve over time. I’d love feedback on the rule set especially: what checks feel useful, what feels too strict, and what’s missing for your own repos.

Report

5mo ago

Maker

Addy Osmani’s recent post on the new software lifecycle has a framing that clicked for me: coding agents are not just the model, they are the model plus the harness around it.

That feels very true when using Claude Code in real repos.

A lot of failures I see are not really “Claude is bad at this” failures. They are repo-environment failures:

no clear agent instructions
unclear conventions
no obvious test command
stale setup docs
unpinned tools
unsafe configs
CI that does not give the agent a clean verification path
tooling that can shift underneath the workflow

In other words, the model might be fine, but the repo gives it a weak harness.

For context, the Addy post I’m referring to is “The New Software Lifecycle.”

I also wrote a related note on the repo-readiness angle here:
https://use-charter.dev/blog/the-harness-is-mostly-your-repo/

Report

5mo ago

Reviews

Maker

📌

Report

5mo ago

Maker

Addy Osmani’s recent post on the new software lifecycle has a framing that clicked for me: coding agents are not just the model, they are the model plus the harness around it.

That feels very true when using Claude Code in real repos.

A lot of failures I see are not really “Claude is bad at this” failures. They are repo-environment failures:

no clear agent instructions
unclear conventions
no obvious test command
stale setup docs
unpinned tools
unsafe configs
CI that does not give the agent a clean verification path
tooling that can shift underneath the workflow

In other words, the model might be fine, but the repo gives it a weak harness.

For context, the Addy post I’m referring to is “The New Software Lifecycle.”

I also wrote a related note on the repo-readiness angle here:
https://use-charter.dev/blog/the-harness-is-mostly-your-repo/

Report

5mo ago