Borrowing it
Nothing to install: this file belongs to joshrotenberg/mcp-proxy. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/joshrotenberg/mcp-proxy/main/AGENTS.mdgit clone --depth 1 https://github.com/joshrotenberg/mcp-proxyWrote 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/joshrotenberg/mcp-proxy/agents-md)<a href="https://agentmods.dev/instructions/joshrotenberg/mcp-proxy/agents-md"><img src="https://agentmods.dev/badge/instructions/joshrotenberg/mcp-proxy/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/joshrotenberg/mcp-proxy/agents-md"><img src="https://agentmods.dev/badge/instructions/joshrotenberg/mcp-proxy/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.04729 | $0.04729 |
| Opus 5.5 | $0.01892 | $0.01892 |
| Sonnet 5.5 | $0.00946 | $0.00946 |
| Haiku 4.5 | $0.00473 | $0.00473 |
Grade A, and why
mcp-proxy 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 17d 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 — 382 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Overview
mcp-proxy is a config-driven MCP (Model Context Protocol) reverse proxy built in Rust. It aggregates multiple MCP backends behind a single HTTP endpoint with per-backend middleware, authentication, and observability. Built on tower-mcp and the tower middleware ecosystem.
Package name: mcp-proxy. Binary name: mcp-proxy. Library name: mcp_proxy.
Project structure
src/
main.rs # CLI entry point (clap), logging setup
lib.rs # Library root, re-exports Proxy and ProxyConfig
proxy.rs # Core: builds proxy, middleware stack, and axum router
lifecycle.rs # Shared runtime owner, scoped transport cancellation and cleanup leases
server.rs # Owned HTTP1/HTTP2 connection/stream tasks and bounded drain
config.rs # TOML/YAML config parsing, validation, all config types
builder.rs # Fluent programmatic configuration builder
discovery.rs # Search index and complete tool schema store
mcp_json.rs # Import Claude-style .mcp.json configurations
ws_transport.rs # Outbound WebSocket transport
skills.rs # Management prompts bundled from src/skills/
admin.rs # HTTP admin API (/admin/backends, /admin/health, etc.)
admin_tools.rs # MCP admin tools (proxy/ namespace, via ChannelTransport)
reload.rs # Owned file watcher and admitted plain backend connections
reload_policy.rs # Frozen startup policy admission and safe last-attempt status
test_util.rs # MockService, ErrorMockService, call_service helper
# Global middleware (wraps the entire proxy, applied in proxy.rs)
access_log.rs # Structured access logging (AccessLogService, target: mcp::access)
audit.rs # Audit completion metadata/error codes, excluding private bodies
alias.rs # Tool renaming (AliasService)
filter.rs # Capability filtering -- allow/deny lists, glob patterns (CapabilityFilterService)
inject.rs # Argument injection into tool calls (InjectArgsService)
mirror.rs # Traffic mirroring to canary backends (MirrorService)
cache.rs # Scoped response caching, shared budgets and generation ownership
cache/ # Bounded SQLite blocking I/O and Redis indexed storage
coalesce.rs # Request deduplication (CoalesceService)
validation.rs # Argument size limits (ValidationService)
metrics.rs # Scoped/shared Prometheus ownership, bounded namespace attribution
telemetry.rs # Privacy-preserving MCP/HTTP spans and explicit OTLP provider factory
rbac.rs # Role-based access control (RbacService)
token.rs # Legacy identity-only helper; no credential forwarding
management.rs # Explicit HTTP/MCP management authority and credential-bound sessions
invocation.rs # Search wrapper resolution preserving caller/request context
bearer_scope.rs # Per-token tool scoping and claims
introspection.rs # OAuth token introspection and JWT fallback
param_override.rs # Tool parameter hiding, renaming, and defaults
composite.rs # Composite tool fan-out
canary.rs # Weighted canary routing
failover.rs # Ordered backend failover
# Per-backend middleware (applied per-backend in proxy.rs and reload.rs)
outcome.rs # Shared response classification, cache admission, hedge adapter
retry.rs # tower-resilience retry layer, exponential backoff, budgets
outlier.rs # Outlier detection and ejection (OutlierDetectionLayer)
tests/
integration.rs # Middleware composition tests with in-process backends via ChannelTransport
e2e.rs # Full proxy pipeline tests with in-process backends
http_e2e.rs # Real HTTP transport and authentication tests
configured_sessions.rs # Configured embedding/CLI full-session transport and policy matrix
support/ # Reusable bounded session/application/backend/OAuth fixtures
health_probes.rs # Configured proxy process probes and auth boundaries
replica_sessions.rs # Two real CLIs through NGINX affinity, retirement/loss and reinitialize
reload_runtime.rs # Watched-file CLI add/replace tests and watcher ownership
coalesce_runtime.rs # Configured HTTP/MCP capacity, expiry and recovery
outcome_runtime.rs # Configured cache recovery, breaker state, hedge winners and metrics
retry_runtime.rs # Configured authenticated sessions and exact volume-budget attempt counts
outlier_runtime.rs # Competing configured calls and raw backend removal/re-registration
shutdown_runtime.rs # SIGTERM/SSE, short drain, construction/drop and stdio reap regressions
cli_logging.rs # Real CLI JSON/text/OTel request records and privacy regressions
outbound_headers.rs # Actual HTTP requests and WebSocket handshake headers
discovery_schemas.rs # Tool schema retrieval and reindex behavior
chaos.rs # Fault injection and resilience recovery regressions
cache_backends.rs # Real Redis containers and temporary SQLite databases
config_properties.rs # Property tests for config validation
invariant_properties.rs # Config/authorization/transform properties and paused-clock schedules
protocol_contracts.rs # Configured library/CLI legacy baseline and modern feature contracts
examples.rs # Load and validate every example TOML config
examples/
*.toml # Example proxy configs for different deployment patterns
docker-compose/ # Docker compose example with HTTP backend
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.
- 17d ago Changed · +86 lines · +1,273 tokens per session 92ae585808a3
- 18d ago Changed · +65 lines · +899 tokens per session 5612d58999dc
- 19d ago Changed · +19 lines · +229 tokens per session 60f92457c760
- 1mo ago Changed · +38 lines · +523 tokens per session 226af6dd00f1
- 2mo ago First seen · 174 lines · 1,805 tokens per session scan A a186886c1c98
mcp-proxy AGENTS.md is an instructions file published in the GitHub repository joshrotenberg/mcp-proxy (19 stars, last pushed today), licensed Apache-2.0. It adds 4,729 tokens to every session, about $0.0189 per session on Opus 5.5. 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.
mcp-proxy AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| mcp-proxy AGENTS.mdjoshrotenberg/mcp-proxy | Instructions file | 19 | today | Apache-2.0 | A | 4,729 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 5,516 | GitHub Copilot |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | today | MIT | A | 4,680 | Codex, OpenCode |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 275k | 3d ago | MIT | A | 4,313 | Codex, OpenCode |