Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/bunizao/edstem-mcp/agents-mdgit clone --depth 1 https://github.com/bunizao/edstem-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/bunizao/edstem-mcp/agents-md)<a href="https://agentmods.dev/instructions/bunizao/edstem-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/bunizao/edstem-mcp/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.02362 | $0.02362 |
| Opus 5 | $0.01181 | $0.01181 |
| Sonnet 5 | $0.00472 | $0.00472 |
| Haiku 4.5 | $0.00236 | $0.00236 |
Grade A, and why
edstem-mcp AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 203 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Operating manual for AI coding agents (Claude Code, Codex, Cursor, etc.) and new human contributors working on this repository.
Read this before making changes. It describes what this project is, how to run and test it, the conventions to follow, and the things never to do.
What this project is
edstem-mcp is a remote Model Context Protocol server that exposes the Ed Discussion (edstem.org) API as MCP tools, so any MCP-compatible client (claude.ai connector, Cursor, Claude Desktop, custom SDKs) can query and interact with EdStem on a user's behalf.
It is the network-deployed sibling of edstem-cli (a local Python CLI). The two are independent codebases; nothing is imported across them. When in doubt about EdStem API behaviour, consult edstem-cli/edstem_cli/client.py as the reference implementation.
The end-to-end implementation roadmap is in PLAN.md. Read that for the "where this is going."
Tech stack
- Runtime: Node.js >= 20
- Language: TypeScript (strict mode), ESM
- HTTP: Express 4
- MCP SDK:
@modelcontextprotocol/sdk(^1.17) - Schema validation:
zod - Tests:
vitest - Dev runner:
tsx watch - Database (planned, M1+): SQLite via
better-sqlite3
No ORM. No framework beyond Express. Keep the dependency surface small.
Setup and common commands
npm install
# Required: an Ed API token from https://edstem.org/settings/api-tokens
export ED_API_TOKEN="your-ed-token"
npm run dev # tsx watch on src/index.ts
npm run check # tsc --noEmit, type-check only
npm run build # compile to dist/
npm run start # node dist/index.js (requires build first)
npm test # vitest run
Server listens on http://localhost:8787/mcp by default.
Environment variables are documented in README.md and (for the planned multi-user mode) in PLAN.md §10.3.
Repository layout
src/
├─ index.ts # Express bootstrap, route mounting
├─ config.ts # env parsing and validation
├─ ed/ # EdStem API client and domain types
│ ├─ client.ts # HTTP client; Bearer auth; error mapping
│ ├─ models.ts # TS types matching Ed JSON
│ ├─ serialization.ts # to-MCP-payload converters
│ └─ filter.ts # post-fetch filtering helpers
├─ mcp/
│ └─ server.ts # MCP tool registry; one server per request
├─ oauth/ # OAuth 2.1 + PKCE provider for MCP auth
│ ├─ provider.ts # EdstemOAuthProvider implementing OAuthServerProvider
│ ├─ store.ts # FileOAuthStore (to be replaced by SQL store in M1)
│ ├─ session.ts # signed cookie session for /authorize page
│ └─ login-page.ts # HTML form rendering
test/ # mirrors src/
dist/ # build output (gitignored)
.data/ # runtime state (gitignored)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 203 lines · 2,362 tokens per session scan A df97db8b0484
edstem-mcp AGENTS.md is an instructions file published in the GitHub repository bunizao/edstem-mcp (0 stars, last pushed 1mo ago), licensed MIT. It adds 2,362 tokens to every session, about $0.0118 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
calc-mcp CLAUDE.md
Claude Code instructions for coo-quack/calc-mcp, covering project rules, tech stack, commands, project structure and tool architecture.
rosetta CLAUDE.md
Instructions for tikoci/rosetta, covering rosetta, project documentation, where does this go?, instruction routing and fast pointers.
rosetta extractor-import-side-effects.instructions.md
Every extractor entrypoint must run its side effects only under import.meta.main. Top-level db.ts imports still open a real DB connection, so tests must import extractors safely.
rosetta database.instructions.md
Use when working on SQLite schema, database initialization, WAL mode, FTS5 triggers, or table definitions.
rosetta issue-pr-linking.instructions.md
Every PR that implements an issue must auto-close it via 'Closes #N'; partial landings spawn follow-up issues before merge.
rosetta manual-version-bump.instructions.md
Version bumps and CHANGELOG promotion are a manual, human step for every release channel. CI must not reintroduce an auto-bump/auto-commit job.