Borrowing it
Nothing to install: this file belongs to HuskyDS/basecamp-mcp-server. 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/HuskyDS/basecamp-mcp-server/main/AGENTS.mdgit clone --depth 1 https://github.com/HuskyDS/basecamp-mcp-serverWrote 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/huskyds/basecamp-mcp-server/agents-md)<a href="https://agentmods.dev/instructions/huskyds/basecamp-mcp-server/agents-md"><img src="https://agentmods.dev/badge/instructions/huskyds/basecamp-mcp-server/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/huskyds/basecamp-mcp-server/agents-md"><img src="https://agentmods.dev/badge/instructions/huskyds/basecamp-mcp-server/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 | $0.06205 | $0.06205 |
| Opus 5.5 | $0.02482 | $0.02482 |
| Sonnet 5.5 | $0.01241 | $0.01241 |
| Haiku 4.5 | $0.00620 | $0.00620 |
Grade A, and why
basecamp-mcp-server 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 2mo 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 — 366 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — working in this repo
Instructions for coding agents (and humans) working on basecamp-mcp-server.
End-user documentation lives in README.md; this file is about how
the code is built.
1. Ground rules
- Never hand-edit
uv.lock. Runuv lock. CI fails on a stale lockfile. - Never add a dependency without stating why in the PR description and in
pyproject.tomlas a comment. The install footprint of a stdio server is a user-visible cost —uvxresolves it on every cold start. src/layout, absolute imports only (from basecamp_mcp.x import y).- Every change updates
CHANGELOG.mdunder## [Unreleased]. - Never log or print a token. See §6.
2. Verified upstream facts
Both dependencies shipped within two weeks of this project starting, and both contain behaviours that look like bugs until you read the source. Every claim below was verified against the installed version and cites where.
Do not "fix" code that contradicts your intuition here without re-verifying first. Several of these are traps that read as obvious mistakes.
mcp 2.0.0
| Fact | Where | Why it matters |
|---|---|---|
FastMCP is now MCPServer; mcp.server.fastmcp does not exist and has no shim |
mcp/server/mcpserver/server.py |
Any snippet you find online from 2025 will not import |
tools/call is not a cacheable method |
mcp_types/methods.py:411 — CacheableMethod lists only prompts/list, resources/list, resources/read, resources/templates/list, server/discover, tools/list |
ttlMs/cacheScope never appear on tool output. All data caching is ours. Locked by test_tools_call_is_not_a_cacheable_method |
A tool may return CallToolResult directly and it passes through untouched |
mcp/server/mcpserver/utilities/func_metadata.py:126 |
The only way to return a structured error. A raised exception is wrapped as a bare string (tools/base.py), losing all structure |
CallToolResult may not appear inside a Union/Optional return annotation |
func_metadata.py:319-323, explicit ValueError |
Return it with no return annotation so output_schema stays None |
call_tool() returns CallToolResult | InputRequiredResult |
func_metadata.py:110 |
Tests must narrow — use tests.conftest.as_tool_result |
ToolAnnotations fields are snake_case: read_only_hint, destructive_hint, idempotent_hint, open_world_hint |
mcp_types |
The wire format is camelCase; the Python API is not |
CacheHint lives in mcp.server.caching, not mcp.types |
mcp/server/caching.py:23 |
Fields are ttl_ms and scope; scope already defaults to "private" |
Cache hints are stored on the lowlevel server as server._lowlevel_server.cache_hints |
mcp/server/lowlevel/server.py:425 |
No public accessor. test_the_hint_is_actually_wired_into_the_server reads the private path deliberately |
tools/list returns tools in registration order (plain dict insertion order) |
tool_manager.py::list_tools |
Never register by iterating a set, a glob, or import side effects. The spec asks for deterministic order because it improves LLM prompt-cache hits |
The SDK serves 2026-07-28 and handshake-era clients simultaneously, routed by the MCP-Protocol-Version header. Nothing is opt-in |
SDK release notes | No shipped client speaks 2026-07-28 yet (Claude connectors top out at 2025-11-25), so ~100% of real traffic takes the legacy leg. Design and test for it |
stateless_http=True is a legacy-leg knob, unrelated to the 2026-07-28 "stateless protocol" |
SDK docs | Do not conflate them |
Roots, Sampling, and protocol-level Logging are deprecated and emit MCPDeprecationWarning |
SDK | filterwarnings = ["error", ...] in pyproject.toml turns accidental use into a test failure |
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.
- 2mo ago First seen · 366 lines · 6,205 tokens per session scan A 0b6c46b68647
basecamp-mcp-server AGENTS.md is an instructions file published in the GitHub repository HuskyDS/basecamp-mcp-server (0 stars, last pushed 3mo ago), licensed MIT. It adds 6,205 tokens to every session, about $0.0248 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.
basecamp-mcp-server AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| basecamp-mcp-server AGENTS.mdHuskyDS/basecamp-mcp-server | Instructions file | 0 | 3mo ago | MIT | A | 6,205 | 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 | 5,685 | GitHub Copilot |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| vscode model-prompts.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 2,022 | GitHub Copilot |