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/airshelf/mcpfs/agents-mdgit clone --depth 1 https://github.com/airshelf/mcpfsWrote 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/airshelf/mcpfs/agents-md)<a href="https://agentmods.dev/instructions/airshelf/mcpfs/agents-md"><img src="https://agentmods.dev/badge/instructions/airshelf/mcpfs/agents-md.svg" alt="Measured on agentmods" 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 | $0.00634 | $0.00634 |
| Opus 5 | $0.00317 | $0.00317 |
| Sonnet 5 | $0.00127 | $0.00127 |
| Haiku 4.5 | $0.00063 | $0.00063 |
Grade A, and why
mcpfs 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 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.
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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Using mcpfs with AI agents
mcpfs mounts SaaS APIs as a FUSE filesystem. Reads via cat | jq, writes via CLI.
Setup
# Install
go install github.com/airshelf/mcpfs/cmd/mcpfs@latest
go install github.com/airshelf/mcpfs/servers/github@latest
# (repeat for: posthog, stripe, vercel, linear, docker, k8s, npm, slack, notion, postgres)
# Mount
source bin/mcpfs-env # load auth tokens
bin/mcpfs-mount # mount all servers at /mnt/mcpfs/
Quick reference — add this to your agent's context
# mcpfs — filesystem reads, CLI writes
## Reads (FUSE filesystem)
cat /mnt/mcpfs/posthog/dashboards.json | jq '.[].name'
cat /mnt/mcpfs/stripe/balance.json | jq '.available[].amount'
cat /mnt/mcpfs/github/repos.json | jq '.[].full_name'
cat /mnt/mcpfs/linear/issues.json | jq '.[].title'
cat /mnt/mcpfs/vercel/deployments.json | jq '.[].url'
cat /mnt/mcpfs/docker/containers.json | jq '.[].Names'
ls /mnt/mcpfs/ # discover all mounted services
## Writes (CLI proxy)
mcpfs-posthog --help # list 67 tools
mcpfs-github --help # list 43 tools
mcpfs-stripe --help # list 28 tools
mcpfs-posthog create-feature-flag --help # per-tool help
## Rules
- PREFER filesystem reads over MCP tool calls (zero token overhead)
- Use jq to filter/transform (universal query language)
- Use native CLIs for writes when available (gh, kubectl, docker)
- Use mcpfs-* CLI for writes when no native CLI exists (posthog, linear)
Auth
All auth via environment variables:
| Server | Env var |
|---|---|
| posthog | POSTHOG_API_KEY |
| github | GITHUB_TOKEN |
| stripe | STRIPE_API_KEY |
| vercel | VERCEL_TOKEN |
| linear | LINEAR_API_KEY |
| notion | NOTION_TOKEN |
| docker | Docker socket |
| k8s | KUBECONFIG |
| postgres | DATABASE_URL |
| slack | SLACK_TOKEN |
| npm | (none) |
Why filesystem over MCP tools?
lscosts 0 tokens. MCP tool schemas cost 20,000+ tokens.cat | jqis one universal query language vs per-tool parameters.- Unix pipes compose across services:
comm -23 <(stripe) <(posthog). - Agents already know
cat,jq,ls. No new abstractions to learn.
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 · 69 lines · 634 tokens per session scan A 8897ec39a5d6
mcpfs AGENTS.md is an instructions file published in the GitHub repository airshelf/mcpfs (4 stars, last pushed 6mo ago), licensed MIT. It adds 634 tokens to every session, about $0.0032 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.
Other instructions, from other repositories
bb AGENTS.md
AGENTS.md instructions for get-bb/bb, covering codebase guidelines, simplicity first, code comments, types and contracts and server and daemon.
ducklab AGENTS.md
Instructions for jrullan/ducklab, covering working in this repository as an agent and path 1: conventional (edit the tree directly).
electron-mcp-server CLAUDE.md
Instructions for halilural/electron-mcp-server, covering project guidelines, git, development and releases.
workz CLAUDE.md
Instructions for rohansx/workz, covering workz mcp integration, setup, available tools and parallel agent workflow.
AvaKit CLAUDE.md
Instructions for mericcintosun/AvaKit, covering claude.md — avakit, what this is, binding rules (do not violate), tech decisions (see docs/04-adr.md) and guardrails.
sandy CLAUDE.md
Instructions for jamestelfer/sandy, covering claude.md, voice, what is sandy, project layout and dev commands.