ctxlayer CLAUDE.md

Developer instructions for ctxlayer, a Cloudflare-hosted service that supplies curated documents to AI tools and connects users to other MCP servers.

In plain words
What is it for?
They help developers navigate and change authentication, document search, MCP connections, REST endpoints, collaborative Markdown editing, and usage administration.
Why use it?
They explain which parts run in the Cloudflare Worker, how the React web app and APIs fit in, and where the architecture documentation lives.

Instructions file

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/stevenn/ctxlayer/claude-md
Clone the repo
git clone --depth 1 https://github.com/stevenn/ctxlayer
Per session 5,018 This file is loaded in full into every session.
When invoked 5,018 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 $0.05018 $0.05018
Opus 5 $0.02509 $0.02509
Sonnet 5 $0.01004 $0.01004
Haiku 4.5 $0.00502 $0.00502

Measured 2d ago against content hash 1a25f69ee02d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ctxlayer CLAUDE.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 2d 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.

CLAUDE.md · 331 lines

How it starts

The opening of the file, as written. The whole thing — 331 lines — stays where its author put it; the contents beside it link to each section on GitHub.

ctxlayer — Claude Code Briefing

This repo is the agent context layer: a remote MCP server on Cloudflare that serves curated docs (with RAG over Vectorize), proxies upstream MCP servers with centralised per-user credentials, and exposes a React SPA for self onboarding + collaborative markdown editing + admin/usage analytics.

docs/PLAN.md is the architecture & data-model reference — not a roadmap. The milestone-driven plan that built ctxlayer (M1–M8) is retired; future work proceeds ad hoc, tracked in code + commits + this file, not in a larger plan. Topic deep-dives are under docs/plan/ (A: auth, B: stdio bridge, C: upstream proxy, D: UI+REST, E: dev environment, F: org IA, G: conventions, I: upstream resilience, M: OKF interop, N: OKF bundles). Skim PLAN.md for the lay of the land and pull in a deep-dive when the topic comes up — but trust the code first; these docs are reference, not kept in lockstep with every change.

What runs where

  • Cloudflare Worker (apps/worker/) is the single deployable unit. It hosts: the MCP server (/mcp, /sse), the OAuth provider (/oauth/*, /.well-known/oauth-authorization-server), REST endpoints for the SPA (/api/*), IdP callbacks (/idp/*), the realtime collab endpoint (/collab/:id), and the React SPA via Workers Assets.
  • React SPA (apps/web/) is built with Vite and shipped from apps/web/dist by the Worker.
  • Shared types (packages/shared/) — Zod schemas + types used by both Worker and SPA.

Architecture pointers

  • MCP per-session state → Durable Object McpSessionDO (apps/worker/src/mcp/session-do.ts).
  • Per-doc realtime collab → Durable Object DocRoomDO (apps/worker/src/collab/doc-room-do.ts).
  • HTTP/SSE upstreams are proxied directly through the generic UpstreamClient interface (apps/worker/src/upstream/*.ts). Stdio MCP servers are not run by ctxlayer: the operator fronts them with their own stdio↔HTTP bridge (e.g. supergateway) and registers the resulting HTTP URL as an ordinary streamable_http upstream (bring-your-own-bridge).
  • All sensitive material is sealed with AES-GCM via crypto/aead.ts.

Read the full file on GitHub · 331 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. 2d ago First seen · 331 lines · 5,018 tokens per session scan A 1a25f69ee02d

Subscribe to this mod's changes

ctxlayer CLAUDE.md is an instructions file published in the GitHub repository stevenn/ctxlayer (9 stars, last pushed 4d ago), licensed MIT. It adds 5,018 tokens to every session, about $0.0251 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.