transcripts-mcp
AGENTS.md8 starsChanged 2 months agoMITView on GitHub
This file provides guidance for AI coding agents (Claude Code, Cursor Agent, Copilot Workspace, Windsurf, etc.) operating on this codebase. transcripts-mcp is a TypeScript pnpm + Turborepo for a local stdio MCP server that queries agent session transcripts. Key directories: Before the human submits your work, ensure:
- Installs packages
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository33 other files it gives its agents.
- .deepsec/AGENTS.mdAGENTS.md
- .cursor/rules/010-workflow.mdcCursor rule
- .cursor/rules/020-codebase-structure.mdcCursor rule
- .cursor/rules/030-coding-style.mdcCursor rule
- .cursor/rules/031-never-use-any.mdcCursor rule
- .cursor/rules/032-import-rules.mdcCursor rule
- .cursor/rules/050-testing.mdcCursor rule
- .cursor/rules/070-comments-defensive-programming.mdcCursor rule
- .cursor/rules/080-ai-interaction.mdcCursor rule
- .cursor/rules/090-changesets.mdcCursor rule
- .cursor/rules/095-calldiff.mdcCursor rule
- .cursor/rules/990-cursor-rule-editing-workaround.mdcCursor rule
- .cursor/rules/999-mdc-format.mdcCursor rule
- adversarial-reviewSkill
- agent-browserSkill
- calldiffSkill
- changesetSkill
- commitSkill
- deslopSkill
- fix-prSkill
- git-conventionsSkill
- humanizerSkill
- improve-codebase-architectureSkill
- logging-best-practicesSkill
- opensrcSkill
- rebaseSkill
- recomposing-commitsSkill
- security-reviewSkill
- show-meSkill
- simplifySkill
- thermo-nuclear-code-quality-reviewSkill
- wait-whatSkill
- writing-for-agentsSkill
The file
We split this AGENTS.md into the rules it gives an agent
10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 7
Check the .cursor/rules directory for relevant guidance before starting any task.
- Distinctive: line 8
Read CONTRIBUTING.md and understand the project's conventions.
- Distinctive: line 9
Read the AI Policy — contributions must comply with it.
- Distinctive: line 34
Write tests for new and existing functionality
- Very specific: line 35
Use oxfmt for formatting and oxlint for linting (not ESLint/Prettier/Biome)
- Very specific: line 36
Follow the existing commit convention: type(scope): description (see .agents/skills/git-conventions/SKILL.md).
- Distinctive: line 37
Before pushing, fetch the upstream branch.
- Very specific: line 41
When a plan, design doc, or PR description covers a real call-flow change that a tree would help reviewers understand, optionally include a …
- Distinctive: line 48
Add new dependencies without justification
- Distinctive: line 54
Before the human submits your work, ensure:
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Instructions for AI Agents
This file provides guidance for AI coding agents (Claude Code, Cursor Agent, Copilot Workspace, Windsurf, etc.) operating on this codebase.
Before You Begin
- Check the
.cursor/rulesdirectory for relevant guidance before starting any task. - Read CONTRIBUTING.md and understand the project's conventions.
- Read the AI Policy — contributions must comply with it.
- Familiarize yourself with the project structure in 020-codebase-structure.mdc.
Project Context
transcripts-mcp is a TypeScript pnpm + Turborepo for a local stdio MCP server that queries agent session transcripts.
Key directories:
apps/mcp— stdio MCP server, tool registration, adapter wiringapps/www— marketing site (Vite + React), deployed with wranglerpackages/core— types, adapter contract, jsonl reader, registrypackages/contracts— serializable MCP tool metadata shared by runtime and docspackages/adapters— Cursor, Claude Code, and Codex harness adapterspackages/search— grep (fff), FTS5, optional semantic searchtools/typescript-config— shared tsconfigtools/oxlint-plugins— custom oxlint rules (anti-slop)distribution/plugin— Cursor Plugin manifest, skill, andmcp.json
Rules for AI Agents
DO
- Help the human contributor understand code and debug issues
- Suggest improvements that follow existing patterns in the codebase
- Write tests for new and existing functionality
- Use oxfmt for formatting and oxlint for linting (not ESLint/Prettier/Biome)
- Follow the existing commit convention:
type(scope): description(see .agents/skills/git-conventions/SKILL.md); commit on the current branch unless the user asks to use a feature branch - Before pushing, fetch the upstream branch. If it has advanced, rebase unpublished local commits onto it, including when working on
main, then push normally. Keep history linear; do not create merge commits or bypass branch protection to push. Rewriting already-published history requires explicit user authorization. - Respect TypeScript strict mode — never use
anyorunknown, use proper types (see 031-never-use-any.mdc) - Cross-package imports use the workspace package name (
@transcripts-mcp/*), never relative paths - Tests are
*.spec.tsundersrc/tests/, never__tests__or*.test.ts - When a plan, design doc, or PR description covers a real call-flow change that a tree would help reviewers understand, optionally include a
npx calldiff@latest diff(see 095-calldiff.mdc). Do not add call-stack diffs to routine chat summaries.
DON'T
- Generate entire pull requests autonomously
- Open issues based on static analysis without human verification
- Refactor code without prior discussion with maintainers
- Add new dependencies without justification
- Introduce patterns that don't already exist in the codebase without discussion
- Submit code that the human operator cannot explain or modify
Code Quality Checklist
Before the human submits your work, ensure:
pnpm lint:fixandpnpm formathave been runpnpm knippassespnpm check-typespasses- Changes are tested (manually at minimum, automated tests preferred)
- PR description explains what and why, not just how
- AI usage is disclosed per the AI Policy
Tech Stack Quick Reference
| Layer | Technology |
| Runtime | Bun |
| Language | TypeScript |
| MCP server | apps/mcp (stdio) |
| Package Manager | pnpm (monorepo) |
| Build System | Turborepo |
| Linter/Formatter | oxlint + oxfmt |
| Tests | Vitest (*.spec.ts under src/tests/) |
| Search | fff, FTS5, optional semantic |
Commands
pnpm install
pnpm dev # start apps via turbo
pnpm check-types
pnpm lint
pnpm format
pnpm knip
pnpm testCursor Cloud specific instructions
The repository pins Node 26.8.1 in .nvmrc and pnpm 12.3.3 in package.json's
packageManager. CI uses the Bun minimum from package.json's engines.bun, currently 1.4.0.
Cloud images may expose a different Node version first on PATH; install or activate the
repository pin before running project commands. The install step runs
pnpm install --frozen-lockfile after activation.
Cloud Agent tool-call shells are non-interactive and do not source ~/.bashrc, and the
exec-daemon prepends its own Node 22 to PATH. A bare node therefore resolves to Node 22,
which breaks pnpm lint (oxlint loads the @transcripts-mcp/oxlint-plugins TypeScript plugin,
which needs Node 26 type-stripping). Activate the project toolchain before running any pnpm
command:
export NVM_DIR="$HOME/.nvm"; . "$NVM_DIR/nvm.sh"; nvm install >/dev/null; nvm use >/dev/null
corepack prepare pnpm@12.3.3 --activate
export PATH="$(dirname "$(nvm which 26.8.1)"):$HOME/.bun/bin:$PATH"With that preamble, pnpm install, pnpm check-types, pnpm test, pnpm lint, and
bun apps/mcp/src/index.ts run against the checked Node and pnpm pins. Ensure Bun 1.4.0 is
installed when the image's bundled Bun differs.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

