Precept reads every pull request against the rules your company runs on. Logic bugs get caught in review, not in production.
Free for up to 3 developers · No credit card · GitHub and GiteaGeneric tools check syntax and style. Your incidents come from broken business rules that only your team knows.
Catch business logic bugs before they reach production, without changing how your team works.
Developers open pull requests as usual. Precept reads each one against your rules and standards.
Each finding cites the rule it breaks, with evidence, a confidence level and a suggested fix.
GitHub and Gitea, cloud or self-hosted. Rules sync from the Markdown already in your repositories. Ask about findings from Claude Code or Cursor through MCP.
Track findings by severity and repository, and how many your team resolves before code ships.
Install the GitHub app or point Precept at your Gitea instance. Pick the repositories to review.
Import the Markdown you already have, or write rules in plain language. Scope them to repositories and paths.
Findings appear inline within minutes, each with the rule it cites, a confidence level and a suggested fix.
Business rules, architecture decisions and standards, kept like folders your team already understands. Open one to see what Precept reads.
The same reviews and knowledge reach your editor through MCP, and your clients through docs generated from the code.
Precept is an MCP server. Connect your AI coding tool once and ask what is blocking a pull request, why a rule applies, or whether your last push fixed it. Read only by default, with your own permissions.
Generate documentation from a repository, keep it private or share it with client accounts. When a client has a question, they ask the docs and get an answer with its sources, not a ticket.
Per developer. Start free and upgrade when your team does. Bring your own Gemini, OpenAI or Anthropic key.
Most reviewers look at the code in isolation. Precept also reads your business rules, architecture decisions and standards, and cites the exact rule behind each finding, so you can verify it in seconds.
No. The first reviews work without any rules. Add rules as you find them, or import the Markdown documents already in your repositories.
Yes, if you want it to. Precept sets a commit status you can require in branch protection. You choose the severity that blocks, per repository.
Yes. Connect your instance with an access token and a webhook. Reviews are posted as pull request comments.
Yes. Precept is an MCP server. Each person connects their own tool and gets their own permissions. It is read only unless your organization allows changes such as re-running a review.
Only your team, until you share them. You can invite client accounts, who see nothing else in Precept, or create a public link. Pages marked internal stay hidden from clients and from their AI answers.
Yours. Connect your own API key from Google Gemini, OpenAI or Anthropic. You pay the provider directly, and you can switch at any time. No key yet? Gemini has a free tier.
For each review, Precept sends the diff, nearby code and the relevant rules to the AI provider you chose, using your key. Data handling then follows that provider’s terms. [Confirm Precept’s own retention terms before launch.]
Every finding shows its rule, the evidence behind it and a confidence level. Mark a finding as not a violation and Precept records it, so the team can refine the rule.
Connect a repository, add your first rule and review a real pull request today.