Skip to content

Veridu changelog

io.github.michaelw-pr/veridu

8 changes recorded. MCP Harbor re-reads this listing from the official registry every six hours and asks the live server what it exposes, and keeps what changed between one look and the next.

  • server.json 2 fields changed
    description
    Your agent verifies its change before commit; PR review that knows what depends on your change. Verify AI-generated code before you commit: independent reviewers check it. PR review too.
    version
    1.1.1 1.1.2

  • llms.txt +27 −4

    https://veridu.io/llms.txt

    − > Veridu verifies code that AI coding agents write, before the pull request and on it. Before the PR, the agent calls `verify_change` with its diff, fixes the grounded findings it gets back and calls it again; Veridu links the follow-up and reports what persists, what is new and what is gone. On the pull request, Veridu finds the callers, implementations and references the change affects, has independent reviewers check the change in that context, drops findings it cannot ground in the code, and− - Free tier: 20 reviews a month, no card. Subscribers: Solo and Team plans.− - [How it works](https://veridu.io/about): change-impact evidence (callers you didn't update, implementations that no longer fit, references to removed code), corroboration across independent reviewers, verdicts− - [GitHub App](https://veridu.io/repos): automatic reviews on every pull request+ > Veridu verifies code that AI coding agents write, before the pull request and on it. Before the PR, the agent calls `verify_change` with its diff, fixes the findings that hold up and calls it again; Veridu links the follow-up and reports what persists, what is new and what is no longer reported (which is not proof of a fix). On the pull request, Veridu looks for the code the change affects (callers and overrides are traced for PHP), has independent reviewers check the change in that context, d+ - ChatGPT: install from the plugin directory listing, https://chatgpt.com/plugins/plugin_asdk_app_6ad14d37e2c08191994915f2450a186c+ - Claude (claude.ai and the Claude apps): add from the connector directory listing, https://claude.ai/directory/connectors/veridu+ - VS Code (GitHub Copilot): listed in the GitHub MCP Registry, https://github.com/mcp/michaelw-pr/veridu ; one-click install at https://veridu.io/verify-vscode-changes+ - Free tier: up to 20 reviews a month, no card. Pull request reviews are for public GitHub repos; verify_change checks the code it is sent, from any repo, unless the change is linked to a private repo. Subscribers: Solo and Team plans (private repos).+ - Who pays for the review: Veridu. For hosted reviews, Veridu's reviewers run on Veridu's own OpenAI and Anthropic API access; a new change or pull request head counts against the user's Veridu reviews, not their Claude or Codex plan. The agent's plan covers only the ordinary MCP tool calls that submit a review and read its result. GitHub App pull request reviews need no AI subscription.+ - [Verify Claude Code changes](https://veridu.io/verify-claude-code-changes): add Veridu to Claude Code, the CLAUDE.md line, and the verify, fix, re-verify loop before commit+ - [Verify Codex changes](https://veridu.io/verify-codex-changes): the same loop in the Codex CLI, with the AGENTS.md line+ - [Review AI-generated pull requests](https://veridu.io/review-ai-generated-pull-requests): what to check outside the diff of an agent-written pull request, and how the GitHub App reviews it+ - [Find what a change breaks](https://veridu.io/check-what-a-change-breaks): a checklist for the code a change affects outside the diff, and what Veridu traces per language+ - [AI code review false positives](https://veridu.io/ai-code-review-false-positives): where AI review noise comes from and the rules Veridu uses to decide what can block and what is shown+ - [Verify Cursor changes](https://veridu.io/verify-cursor-changes): add Veridu to Cursor as a remote MCP server (browser sign-in where the client supports it; an MCP-only token is the verified fallback), the project rule line, and the verify, fix, re-verify loop before commit+ - [Verify PhpStorm changes](https://veridu.io/verify-phpstorm-changes): add Veridu to JetBrains AI Assistant, Junie or GitHub Copilot in PhpStorm as a remote MCP server (a personal connection URL from https://veridu.io/connect for AI Assistant; Junie browser sign-in where it works, an MCP-only token otherwise and for Copilot), the guidelines line, and sending the PHP callers and interfaces around a change so callers and overrides can be traced+ - [Verify VS Code changes](https://veridu.io/verify-vscode-changes): install Veridu in VS Code with one click (vscode:mcp/install link) or by hand (.vscode/mcp.json "servers" with type http and the URL), VS Code signs in to Veridu itself, the Copilot instructions line, and the verify, fix, re-verify loop before commit+ - [Find every caller of a function](https://veridu.io/find-callers-before-changing-a-function): what your editor's find references misses, and which callers Veridu traces for you, by language+ - [Automatic pull request review](https://veridu.io/automatic-pull-request-review): install the GitHub App, when a pull request is reviewed, and what blocks versus what is shown+ - [Can one AI review code another AI wrote?](https://veridu.io/can-one-ai-review-code-another-ai-wrote): where a second model falls short, and the rules Veridu uses (independent reviewers from different vendors, grounded findings, a human decides)+ - [Worked example](https://veridu.io/worked-example): one real verify loop on a small PHP change: Veridu returns a block for a missed caller, the agent fixes it, the follow-up shows what is no longer reported+ - [For software agencies](https://veridu.io/for/software-agencies): developers check each change before asking for review; pull request reviews on client repositories; data handling and the GitHub App's permissions+ - [MCP server for code review](https://veridu.io/mcp-code-review): the endpoint, browser sign-in, the four tools (verify_change, get_change, veridu_review, veridu_review_status) with what each takes and returns, and the clients it works in+ - [Code review in ChatGPT](https://veridu.io/code-review-in-chatgpt): add the Veridu app from the ChatGPT directory, paste a pull request link or a diff, and what comes back+ - [Code review in Claude](https://veridu.io/code-review-in-claude): add the Veridu connector from the Claude directory, paste a pull request link or a diff, and what comes back+ - [Security](https://veridu.io/security): the sandbox with no internet access, GitHub permissions, which providers see code, retention, public versus private reviews, and --local mode+ - [Review before a human](https://veridu.io/review-before-a-human): opt-in per repo; Veridu reviews every push to a draft PR, the author fixes or dismisses findings, and a handoff (not an approval) is added when nothing at the display level is left open+ - [How it works](https://veridu.io/about): change-impact evidence (callers you didn't update, implementations that no longer fit, references to removed code; callers and overrides are traced for PHP), corroboration across independent reviewers, verdicts+ - [Privacy policy](https://veridu.io/privacy): personal data, subprocessors and deletion+ - [GitHub App](https://veridu.io/docs#github-app): automatic reviews when a pull request is opened, reopened or marked ready for review

  • server.json 2 fields changed
    repository_url
    — https://github.com/michaelw-pr/veridu-plugin
    version
    1.1.0 1.1.1

  • llms.txt +4 −3

    https://veridu.io/llms.txt

    − > Veridu is code review that knows what depends on your change. It finds the callers, implementations and references a change affects, has independent reviewers from different vendors check the change in that context, drops findings it cannot ground in the code, and blocks only on problems both reviewers confirm. It marks a review incomplete rather than passing code it could not see. Coding agents add it with one command as an MCP server (https://api.veridu.io/mcp, browser sign-in, no API keys) − - [Three AI code reviewers, same 18 PRs: two mostly found a subset of the third](https://veridu.io/blog/three-reviewers-18-prs/): On identical input, findings ranged from 40 to 2, and almost everything the smaller reviewers found, the largest found too. What a second reviewer is for.− - [How it works](https://veridu.io/about): change-impact evidence (callers you didn't update, implementations that no longer fit, references to removed code), cross-vendor reconciliation, verdicts+ > Veridu verifies code that AI coding agents write, before the pull request and on it. Before the PR, the agent calls `verify_change` with its diff, fixes the grounded findings it gets back and calls it again; Veridu links the follow-up and reports what persists, what is new and what is gone. On the pull request, Veridu finds the callers, implementations and references the change affects, has independent reviewers check the change in that context, drops findings it cannot ground in the code, and+ - [How to Find a Pull Request’s Blast Radius](https://veridu.io/blog/pull-request-blast-radius/): A worked review of fastify#6716, with a manual method for finding affected code, documentation, and contracts outside the diff.+ - [Three AI code reviewers, same 18 PRs: two mostly found a subset of the third](https://veridu.io/blog/three-reviewers-18-prs/): On identical input, findings ranged from 40 to 2, and almost everything the smaller reviewers found, the largest found too. What a second reviewer is for.+ - [How it works](https://veridu.io/about): change-impact evidence (callers you didn't update, implementations that no longer fit, references to removed code), corroboration across independent reviewers, verdicts

  • llms.txt +7 −2

    https://veridu.io/llms.txt

    − > Veridu is code review that knows what depends on your change. It finds the callers, implementations and references a change affects, has independent reviewers from different vendors check the change in that context, drops findings it cannot ground in the code, and blocks only on problems both reviewers confirm. It marks a review incomplete rather than passing code it could not see. Coding agents add it with one command as an MCP server (https://api.veridu.io/mcp, browser sign-in, no API keys) − - Free tier: 5 reviews, no card. Subscribers: Solo and Team plans.+ > Veridu is code review that knows what depends on your change. It finds the callers, implementations and references a change affects, has independent reviewers from different vendors check the change in that context, drops findings it cannot ground in the code, and blocks only on problems both reviewers confirm. It marks a review incomplete rather than passing code it could not see. Coding agents add it with one command as an MCP server (https://api.veridu.io/mcp, browser sign-in, no API keys) + - Free tier: 20 reviews a month, no card. Subscribers: Solo and Team plans.+ - [Change impact analysis](https://veridu.io/change-impact-analysis): what Veridu looks for outside the diff, which parts are traced automatically, and two real examples+ - [Laravel code review](https://veridu.io/laravel-code-review): what Veridu traces in PHP, with a real laravel/framework bug it found outside the diff+ ## Blog+ - [Three AI code reviewers, same 18 PRs: two mostly found a subset of the third](https://veridu.io/blog/three-reviewers-18-prs/): On identical input, findings ranged from 40 to 2, and almost everything the smaller reviewers found, the largest found too. What a second reviewer is for.+ - [Our reviewer missed a Laravel bug. Here is what we changed.](https://veridu.io/blog/missed-laravel-bug/): A Laravel bug escaped diff-only review because the broken override was outside the pull request. Here is how we added bounded change-impact context.

  • server.json 2 fields changed
    description
    Code review that knows what depends on your change — before commit or on every PR. Your agent verifies its change before commit; PR review that knows what depends on your change.
    version
    1.0.0 1.1.0

  • llms.txt 1 field changed
    file
    absent published

    https://veridu.io/llms.txt

    + # Veridu+ > Veridu is code review that knows what depends on your change. It finds the callers, implementations and references a change affects, has independent reviewers from different vendors check the change in that context, drops findings it cannot ground in the code, and blocks only on problems both reviewers confirm. It marks a review incomplete rather than passing code it could not see. Coding agents add it with one command as an MCP server (https://api.veridu.io/mcp, browser sign-in, no API keys) + To install Veridu into an AI coding agent, follow https://veridu.io/install.md — it is written for agents to execute.+ - Claude Code: `claude mcp add --transport http veridu https://api.veridu.io/mcp`, then `/mcp` → veridu → Authenticate+ - Codex CLI: `codex mcp add veridu --url https://api.veridu.io/mcp && codex mcp login veridu`+ - Any other MCP client: remote streamable-HTTP server at `https://api.veridu.io/mcp` (OAuth sign-in)+ - Free tier: 5 reviews, no card. Subscribers: Solo and Team plans.+ ## Docs+ - [Install guide for agents](https://veridu.io/install.md): add the MCP server, sign in, verify a change, review a PR, troubleshooting+ - [Documentation](https://veridu.io/docs): MCP, CLI, web and GitHub App surfaces, verdicts, limits+ - [Pricing](https://veridu.io/pricing): free tier and plans+ - [Demo](https://veridu.io/demo): a real two-reviewer review, rendered from stored output+ ## What it checks+ - [How it works](https://veridu.io/about): change-impact evidence (callers you didn't update, implementations that no longer fit, references to removed code), cross-vendor reconciliation, verdicts+ - [Data handling](https://veridu.io/data): what is sent, stored and for how long+ ## Optional+ - [GitHub App](https://veridu.io/repos): automatic reviews on every pull request+ - [Status](https://veridu.io/status)

  • server.json 2 fields changed
    description
    Multi-reviewer AI code review: surfaces findings the reviewers agree on, disputes flagged. Code review that knows what depends on your change — before commit or on every PR.
    version
    0.1.0 1.0.0