ubukhari

The step where your MCP server's OAuth sign-in breaks: walks the 401, RFC 9728 and RFC 8414 discovery and client registration from outside and says what breaks for claude.ai, Claude Code, ChatGPT, VS Code and Cursor, whose it is, and the fix.

1 plugin · ubukhari/mcp-oauth-doctor

Why your MCP App renders blank: checks a remote MCP server's ui:// views (tool link, MIME type, CSP, the ui/initialize bridge, ui.domain) for Claude, ChatGPT and any MCP Apps host, plus ChatGPT's directory review, says whose it is and the fix, and previews the view sandboxed.

1 plugin · ubukhari/mcp-apps-doctor

A readable receipt of what your agent did this week, from your own logs: Claude Cowork, Claude Code and Codex, plus Zapier, Make or n8n run exports, by folder: sends, writes, deletes, approvals, failures, tokens, cost. Local; sends nothing.

1 plugin · ubukhari/agentreceipt

Which merges an AI agent wrote, and whether a person approved them: every merged PR in a period with its agent (Copilot, Claude Code, Codex, Cursor, Devin…), its approvals, who merged it and the SOC 2 CC8.1 exceptions, as a dated CSV + markdown. Not audit advice.

1 plugin · ubukhari/agent-merge-audit

Which model actually answered your coding agent, from your own machine: Claude Code and Codex transcripts read locally for model switches and the notices before them, reasoning-token spikes at fixed values, and context-ceiling changes. Sends nothing.

1 plugin · ubukhari/which-model

Your sub-processor list, drafted from your code: every vendor your app sends data to, with purpose, a data guess, region, DPA and the file and line it was found at. A CLI and a GitHub Action that flags a PR adding a vendor. Not legal advice.

1 plugin · ubukhari/subprocessors

What your coding-agent seats and tokens buy, from your own machine: Claude Code and Codex tokens and API-equivalent cost by day, project and model, burn toward your own limits, and a flag when the rate changes. Local, no account.

1 plugin · ubukhari/seatledger

One dollar ceiling across the coding agents on your machine: set 40/day, /week or /run, measured from Claude Code and Codex transcripts and enforced by pre-tool-use hooks in Claude Code, Codex and Gemini CLI, with a runaway guard. Cursor is advisory. Local.

1 plugin · ubukhari/agent-cap

The records a small employer keeps when AI screens job applicants, before Colorado's SB 26-189 (1 Jan 2027): an inventory of the AI tools in your hiring from your own ATS export, candidate notices, Colorado's 30-day adverse-outcome explanation, review procedure and logs, as dated files. Local. A draft for your counsel to review; not legal advice.

1 plugin · ubukhari/hiringrecord

EU AI Act Article 50 readiness check, from outside: does your chatbot tell people it is an AI, and are your AI images, audio, video and text marked (C2PA, IPTC, XMP)? Quotes the law with its dates; not legal advice. No account, no LLM; files read locally.

1 plugin · ubukhari/agent-disclosure

Find the API versions your code pins (Shopify, Meta Graph and Marketing, Google Ads, HubSpot, Stripe, Twilio Flex, Keap) and when each stops working: days left, what breaks, whether it fails silently, the replacement. From the providers' own pages; no account, no keys.

1 plugin · ubukhari/api-sunset

Find the LLM model IDs in your code that OpenAI, Anthropic or Google is retiring: shutdown date, recommended replacement, its price ratio and the API changes to make, from the providers' own pages. No account, no keys.

1 plugin · ubukhari/model-migrate

R&D tax-credit evidence from your git history: a tax year's commits by component and person, and the commits that look like experimentation (reverts, unmerged branches, spikes, benchmarks), each traced to its commit, in a dated pack for your accountant. Local; no account, no LLM. Not tax advice.

1 plugin · ubukhari/rd-ledger

What your own last 30 days of Claude Code and Codex would cost on every coding plan: Claude Pro, Max and Team, Codex on Plus, Pro and credits, the APIs, and Copilot AI credits as an estimate. Local; sends nothing.

1 plugin · ubukhari/tokens-per-dollar

Did your scheduled GitHub workflows run when they should have? Late, missed, failed and green-but-empty runs, workflows GitHub disabled, and the 60-day countdown, from your repo's own run history with your own token. CLI, GitHub Action, MCP tool.

1 plugin · ubukhari/didrun

Who needs a 1099 for 2025 or 2026 payments, from your own QuickBooks, Xero or vendor-payments CSV: the $600 and $2,000 thresholds, card and PayPal amounts left to the 1099-K, corporations exempt except attorneys and medical, missing W-9s, and IRIS Taxpayer Portal CSVs. Local; TINs never leave the machine.

1 plugin · ubukhari/iris-1099

Nothing an AI writes to QuickBooks or Xero posts until you say yes: a hook for Claude Code, Cowork, Codex and Gemini CLI that holds every write through the QuickBooks and Xero MCP servers and APIs, with a dry-run diff and a per-client log. Local; no account.

1 plugin · ubukhari/booksguard

Every GitHub Actions removal your workflows will hit, with the date: retiring runner labels (macos-14, ubuntu-22.04) and their brownouts, node20 actions, failing artifact and cache versions, set-output. From GitHub's own announcements; a CLI and an Action.

1 plugin · ubukhari/actions-eol

Tie Stripe payouts to your bank deposits, and your QuickBooks or Xero books to your bank statement month by month, from your own CSV exports: timing, fees, refunds and disputes, and what does not tie with the likely reason. Local; no account, no bank linking.

1 plugin · ubukhari/tieout

Nothing irreversible without an OK. okgate (formerly agentguard), the MCP policy proxy for AI agents: hard spend limits, destructive-action approvals, kill switch, scoped keys, dry-run diffs, loop breaker, hash-chained audit log.

2 plugins · ubukhari/okgate

A weekly Postgres backup that is restored and checked every time, in your own GitHub Actions: Supabase (Free has no backups), Neon or any Postgres. pg_dump, a restore into a throwaway copy, every table and row compared. Nothing hosted.

1 plugin · ubukhari/restoreproof