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/rahmanef63/connectors/agents-mdgit clone --depth 1 https://github.com/rahmanef63/connectorsWhat 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 | $0.02558 | $0.02558 |
| Opus 5 | $0.01279 | $0.01279 |
| Sonnet 5 | $0.00512 | $0.00512 |
| Haiku 4.5 | $0.00256 | $0.00256 |
Grade A, and why
connectors AGENTS.md scanned grade A with 1 finding 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 3d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **`support.claude.com`** returns 200 to `curl`, but the body is a JS shell — the article text hydrates client-side, so a plain fetch reads as an empty page rather than as a failure. **A rendering fetch tool does read i How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Scope: the entry point for an agent reading this repo cold — full file map, reading order per goal, and the invariants your output must not break. Assumes: you were pointed here to write or debug an AI integration in another codebase. Every file here is read-only input; nothing in this repo is installed, imported or executed.
Goal
Produce one remote MCP server in the target codebase, plus whatever registration the user's chosen host needs. One server, several ways to register and ship it. Vendor folders describe registration and distribution only — they never change server logic.
Load only what applies
Every Markdown file here is guidance; there are no dated reports to skip. Reading all of them still costs tokens you want for the build. Open the folder that matches the goal, then pull single shared/ files on demand (shared/README.md is that folder's own table). The per-goal orders below are the intended budget.
File map
| Path | Purpose |
|---|---|
README.md |
human router: thesis, folder decision table, dependency line, license |
AGENTS.md |
this file — agent router |
LICENSE |
MIT, 2026 rahmanef63 |
packages/mcp-files/ |
code — the shared OpenAI file-input contract: schema builders, SSRF-safe ingestion, store/attach adapter seams. README.md there is the quickstart |
cn-mcp-core/README.md |
the design invariant, the phase decision tree, framework/DB adaptation notes, spec links |
cn-mcp-core/phase-1-bearer.md |
the endpoint: JSON-RPC dispatch, bearer auth, tool registry |
cn-mcp-core/phase-2-oauth.md |
the authorization half: consent page, auth codes, token exchange, discovery documents |
cn-mcp-core/phase-3-admin-ui.md |
where a human mints, inspects and revokes their own tokens |
shared/README.md |
router for shared/ — the "open when" table for all thirteen files |
shared/tool-design.md |
tool vs resource vs prompt, naming, granularity, annotations — decides whether the server is any good |
shared/results.md |
what tools/call hands back: structuredContent, the outputSchema one-way door, isError |
shared/testing.md |
catalog snapshot, host-facing invariants, golden prompts — the failures that never throw |
shared/clients.md |
per-host matrix: transport accepted, registration path, whether a bridge is needed |
shared/oauth.md |
OAuth 2.1 + PKCE S256 in implementable detail |
shared/transport.md |
POST /mcp JSON-RPC shape, SSE, status codes, discovery documents |
shared/convex.md |
Convex-only gotchas, starting with the SITE vs CLOUD origin split |
shared/setup-form.md |
the copy-paste-ready setup form — the exact string each host wants, the card spec, a dependency-free CopyField |
shared/file-inputs.md |
files and images both directions — ChatGPT's file-param contract, the SSRF and size guards, where the bytes live, ImageContent on the way out |
shared/icons.md |
every image a connector needs — MCP's Icon schema, OpenAI's two required squares, Claude's absent field, one asset set for all three |
shared/pitfalls.md |
16 real failures, symptom → root cause → fix |
shared/security-checklist.md |
pass/fail gate before the endpoint is reachable by anyone |
cn-claude-plugin/README.md |
connector vs plugin, the surface picker, the claude.ai connector path |
cn-claude-plugin/manifest.md |
plugin.json fields, component directories, ${CLAUDE_PLUGIN_ROOT}, local testing |
cn-claude-plugin/marketplace.md |
marketplace.json, /plugin install, what catalog submission actually is |
cn-gpt-plugin/README.md |
the auth-mode gate (no API-key field in ChatGPT) and the Apps SDK → Plugins rename |
cn-gpt-plugin/register.md |
Path A — developer mode, your own or your workspace's ChatGPT, no review |
cn-gpt-plugin/publish.md |
Path B — public Plugins Directory submission, verification and tool scan |
cn-gpts/README.md |
Actions vs MCP decision, plus the eight GPT Builder Configure fields |
cn-gpts/openapi-actions.md |
the OpenAPI 3.1 document, auth panel, privacy-policy gate, when to abandon this for MCP |
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.
- 3d ago First seen · 89 lines · 2,558 tokens per session scan A 091d487981d2
connectors AGENTS.md is an instructions file published in the GitHub repository rahmanef63/connectors (2 stars, last pushed 8d ago), licensed MIT. It adds 2,558 tokens to every session, about $0.0128 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
enquire-mcp AGENTS.md
Instructions for oomkapwn/enquire-mcp, covering agents.md, tl;dr, architecture (5-minute orientation), conventions and typescript.
mcp-app-typescript-template AGENTS.md
AGENTS.md instructions for pomerium/mcp-app-typescript-template, covering repository guidelines, project overview, development commands, quick start and all available commands.
Local-Codex-Bridge AGENTS.md
Instructions for zoeynine/Local-Codex-Bridge, covering ai agent guidance, project intent, architecture, security and permission boundaries and adapting to another machine.
ThreeBrowserStudio AGENTS.md
Instructions for SamG-Coder/ThreeBrowserStudio: ThreeBrowser Studio is an LLM-first authoring runtime. MCP is the editor and the native WebGPU window is the shared live viewport. Preserve that product shape: do not grow a conventional inspector-heavy browser UI.
PatchWarden AGENTS.md
Instructions for jiezeng2004-design/PatchWarden, covering patchwarden agent rules, commands, safety contracts, changes and release and codex memory.
sdk CLAUDE.md
Instructions for WaniWani-AI/sdk, covering waniwani sdk, public api tiers, oss (no api key required), free tier (one env var: waniwaniapikey) and internal (not part of the public api).