vrchat-mcp AGENTS.md

Repository instructions for the VRChat MCP project, which connects an AI agent to tools for VRChat. They describe checks, configuration, development commands, review rules, and logging requirements.

In plain words
What is it for?
Use them when modifying or reviewing BASIC-BIT/vrchat-mcp. They cover linting, type checking, tests, configuration overrides, login, pull-request review, and stderr-only logging.
Why use it?
They give coding agents project-specific rules so changes are tested, configured correctly, and do not corrupt MCP communication on standard output.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/basic-bit/vrchat-mcp/agents-md
Clone the repo
git clone --depth 1 https://github.com/BASIC-BIT/vrchat-mcp

Made for: Codex, OpenCode.

Per session 1,604 This file is loaded in full into every session.
When invoked 1,604 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.01604 $0.01604
Opus 5 $0.00802 $0.00802
Sonnet 5 $0.00321 $0.00321
Haiku 4.5 $0.00160 $0.00160

Measured yesterday against content hash 0303a03fcb13, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vrchat-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 yesterday.

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.

AGENTS.md · 84 lines

How it starts

The opening of the file, as written. The whole thing — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS

This repo uses linting, typechecking, and tests to validate changes.

Guidance for coding agents:

  • After significant code changes (or at the end of a work loop), run npm run check.
  • If you need just one step, use npm run lint, npm run typecheck, or npm test.
  • After making a change, run at least one relevant targeted test (or add/edit one) and confirm it passes before reporting back.
  • For PR review workflow: before pushing any new commits, respond to each open review comment (reply or reaction) and resolve the thread.
  • Claude reviews non-draft, non-fork PRs automatically and rewrites one sticky comment per PR. The review job uses GitHub OIDC to retrieve BASIC's canonical Claude OAuth token from AWS Secrets Manager; there is no repository copy to rotate. Calibration lives in REVIEW.md; add the skip-claude-review label to opt a PR out.
  • Keep stdout reserved for MCP protocol; log to stderr only.
  • Config defaults live in src/config/defaults.json; override via VRCHAT_MCP_CONFIG_FILE (env overrides still supported).
  • Prefer the local harness over Codex MCP hosting for dev iteration:
    • npm run mcp:login to authenticate (keeps the server alive while you log in).
    • npm run mcp:status, npm run mcp:logout, npm run mcp:list-tools, npm run mcp:call.
    • The harness defaults cookie storage to file. Override via config JSON or env vars if needed.
    • Use VRCHAT_MCP_SERVER_COMMAND / VRCHAT_MCP_SERVER_ARGS to switch between tsx src/index.ts and built dist/bin/cli.js.
  • Write tools are enabled by default. Use writes.allow=false in config, or VRCHAT_MCP_ALLOW_WRITES=false, when a test or run must be read-only.
  • Regenerate tool catalog docs after spec updates: npm run generate:tools-docs.
  • Regenerate VRChat OpenAPI schemas after spec updates: npm run generate:schemas (updates src/generated/vrchat-schemas.ts; do not edit manually).
  • Regenerate mock test schemas after spec tweaks: npm run generate:test-schemas (updates test/generated/mock-schemas.ts; do not edit manually).
  • specs/vrchat-openapi.yaml is the community spec, not VRChat's, and drifts from the live API continuously. docs/spec-drift.md logs divergences we have observed so nobody re-investigates them — it is a record, not an authority; the live API is the source of truth. Read it before trusting a spec constraint, re-verify any entry you are about to depend on for a substantive change, and add an entry whenever you find a new divergence. Fix drift in scripts/postprocess-schemas.ts, not in the spec: the spec file is gitignored, so edits there are local-only and vanish on a fresh clone.

Read the full file on GitHub · 84 lines

Changes

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.

  1. yesterday First seen · 84 lines · 1,604 tokens per session scan A 0303a03fcb13

Subscribe to this mod's changes

vrchat-mcp AGENTS.md is an instructions file published in the GitHub repository BASIC-BIT/vrchat-mcp (2 stars, last pushed 8d ago), licensed MIT. It adds 1,604 tokens to every session, about $0.0080 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.