mcptoon

Install 1,000 skills and 1,000 MCP tools locally — and don't worry about the token context. mcptoon manages it all.

Its own compact format does token optimization, tool discovery and context compression at once: it cuts the tool context by ~90% (99.2% on a 255-tool sample; run mcptoon status for your own number) and writes no line of config for any desktop or command-line agent.

Connects to a 17,000+ MCP tool registry and searches skills on demand — nothing pre-installed, you pick what goes in.

It's just a 341KB native CLI — delete it anytime; keep it, and you never have to configure tools or skills for any agent again.

👉 See it first: landing page · 30-second token calculator · 中文

👉 Or run it: pip install mcptoon && mcptoon bench — it measures what your own tools and skills cost, on your machine.

👉 Developer docs · Contributing · Issues

Why mcptoon

Every MCP agent (Claude Code, Cursor, Codex, …) loads the full description of every tool and every skill into its context window before it does any work — and re-sends it every turn. On 255 tools that is 71,929 tokens: more than half of a 128K window spent on descriptions, before the first question.

How the token saving works

mcptoon keeps those descriptions on disk, not in context, and hands the agent a compact view instead:

• A name index, not full schemas. Picking a tool only needs its name — 255 tools become 581 tokens (−99.2%). The full schema is fetched on demand with mcptoon inspect, only when a tool is actually called.
• Skills the same way. One resident pointer (39 tokens) plus one lookup (501 tokens) replaces loading every SKILL.md in full — 926,232 → 39 + 501 (−99.9%).
• Results too. --toon re-encodes a tool's result as standard TOON. Measured 5.7% smaller on a mixed result sample, ranging 1.8%-48.6% by result shape — a list of records barely shrinks, a single record halves. Not a flat 34%: see assets/benchmark_results.json.

Nothing is lost: the full schema and the full skill text stay one command away. Only the context window is spared.

mcptoon is a 341KB, zero-dependency native CLI that manages every MCP tool and agent skill on your computer — and shares them across all your agents with no config written by you.

This isn't just us talking

Those numbers are ours — but "loading every tool schema into context is expensive" is not a claim only we make:

• Anthropic's engineering write-up — tool schemas flooding the context window is a real cost; one example drops from 150,000 tokens to 2,000
• Firecrawl's benchmark — the same task cost 1,365 tokens via CLI vs 44,026 via MCP (32×)
• Scalekit's benchmark — CLI 10–32× cheaper, 100% reliable vs MCP's 72%
• MCP-Zero (arXiv:2506.01056) — on-demand tool retrieval, near-constant cost regardless of tool count

mcptoon is the one you can use today, covering every agent at once.

Two ways in

Measure your numbers: pip install mcptoon && mcptoon bench — it reports your own catalog, not the fixed 255-tool sample below (that sample's method is in docs/tiktoken-benchmarks.md).

Skills work the same way: 926,232 tokens of SKILL.md text → 39 resident + 501 per lookup (−99.9%). Call results shrink a further ~6% with --toon (shape-dependent).

Path 1 · I just use AI tools

You have a desktop AI — Claude, Codex, Cursor, Windsurf, Cline, VS Code Copilot. Today, adding a tool or a skill means hand-editing that agent's JSON. mcptoon removes that step:

1. Install once.

   bash

   pip install mcptoon && mcptoon quickstart
   

   quickstart finds your MCP servers, writes your config, and registers mcptoon in every agent it detects. (Skip quickstart and the first command you run still self-heals: it installs mcptoon's own skill into each agent and builds the skill index, once per machine.)

2. Add tools and skills in one place — here, not in each agent.

   bash

   mcptoon install --search github     # find and install any MCP server
   mcptoon skills search "make a PDF"  # find a skill by what it does
   

3. Use them from any agent. Your agent runs mcptoon like any other command — no config, no restart. See Works with every AI agent.

Want a head start? Four built-in starter packs — essentials, web-research, code-review, docs — stand up a working toolset in one command:

mcptoon install --pack essentials

Path 2 · I build with MCP / agents

mcptoon is a plain CLI with scriptable output, plus an MCP endpoint when you want one.

mcptoon manifest --format json      # machine-readable tool index
mcptoon call <server> <tool> '{}'   # call any tool, JSON or --toon output
mcptoon serve                       # expose every configured server behind one MCP endpoint

• Stable output — JSON by default, --toon for smaller results (shape-dependent, 5.7% on a mixed sample), --format mcp to export standard MCP JSON.
• mcptoon serve — stdio or HTTP, connection pooling, per-agent keys, for clients that insist on a proxy.
• Zero dependencies — pure Python standard library, so it drops into any environment (CI, containers, air-gapped).

Full reference: DEVELOPERS.md and All commands.

Contents

• Why mcptoon
• How the token saving works
• This isn't just us talking
• Two ways in
• 30 seconds up and running
• What it does
• Where the tools come from
• The three bills
• Install
• Works with every AI agent
• Why a CLI, not a proxy
• All commands
• Trust and safety
• Credits and references
• Contributing
• License

30 seconds up and running

pip install mcptoon                          # pure stdlib, 341KB, zero dependencies

# One command: find your MCP servers, write your config, register the gateway
# in every agent you have, and show you what it found:
mcptoon quickstart

# See every tool available (names-only by default; 255 tools cost 581 tokens):
mcptoon manifest

# Call a tool (JSON output by default; add --toon to save more):
mcptoon call everything echo '{"message":"hi"}'

quickstart is also what makes mcptoon visible: it writes mcptoon serve into each agent's config as the reserved server mcptoon, so your agent can see mcptoon itself. Already synced? Re-register with mcptoon sync --self (plain mcptoon sync only writes your servers); check the state any time with mcptoon status.

Two ways in — and you don't pick one. mcptoon is a CLI first: a shell-capable agent runs mcptoon commands, and that is the cheapest path — the schemas never enter context. quickstart also registers mcptoon's own MCP entry (mcptoon serve) with each agent that takes one, so a client that can only see MCP servers still reaches your tools. The MCP entry ships in compact mode: it keeps your upstream tool list out of the model's context and sends it to mcptoon_manifest → inspect → call instead. mcptoon status shows both ledgers.

And it is fully reversible. The install registers the gateway, and by default quickstart also takes over — it routes your existing servers through the gateway and removes their direct entries (without that step your agent sees the same tools twice, once direct and once through the gateway, which costs more, not less). It is a listed, confirmed step, and it is undoable: mcptoon restore drops the gateway entry and puts your servers back exactly where they were. Anything you added to a config afterwards is left untouched. mcptoon off is the lighter switch — it removes only the gateway entry and leaves your servers where they are. Preview either with --dry, and preview the complete removal plan with mcptoon uninstall --dry (it prints exactly what it will remove first).

Don't want to install yet? Watch it work instead (needs Node):

uvx mcptoon demo --quick

It boots the official "everything" reference server, calls one tool, and prints the token math on your screen — no API key, none of your servers, nothing written to disk.

What it does

Tools: schema on demand

The full schema is compressed to a name index. When you need one, mcptoon inspect fetches its real parameters, then call runs it. You pay for a listing, not for every turn.

Skills: one resident pointer

No more loading the whole catalog. A single pointer line stays resident (39 tokens) and one lookup returns the most relevant skills (501 tokens). Views are links (a junction on Windows, no admin needed), so one edit at the source is live everywhere and there is no second copy to drift.

Install once, share everywhere

Install mcptoon once and every AI on the machine gets all your tools and skills. Add more later and it is live immediately — no agent restart, no per-agent JSON.

Add tools your way

mcptoon install brave-search --npm @modelcontextprotocol/server-brave-search
mcptoon install my-tool --pip mcp-my-tool
mcptoon install remote-api --url https://example.com/mcp
mcptoon add my-server --stdio npx -y @any/mcp-package
mcptoon install --list                       # see what's installed
mcptoon install --remove brave-search        # uninstall one

One command per server, from npm / pip / HTTP. Or let mcptoon scan what you already have: mcptoon discover.

A remote server that needs a token reads it from the environment, so no credential is ever written to disk:

mcptoon install remote-api --url https://example.com/mcp \
  --header 'Authorization: Bearer ${BAIZHI_TOKEN}'   # BAIZHI_TOKEN set in your environment

${NAME} is resolved at request time, so the config, the generated handler and every log keep the template — rotate the variable and the next call uses the new value, with no reinstall. A missing or empty variable fails loudly, naming it, never as an empty header.

Where the tools come from — search 17,000+, install with one command

On a new machine the first question isn't "how do I save tokens", it's "which tools do I even want". The old answer is to hunt GitHub for mcpServers snippets and hand-copy them into JSON.

mcptoon ships no tools and bundles no catalog. It queries the upstream registries, so you search for exactly what you need:

mcptoon install --search github     # search, list only — nothing installed
mcptoon install --search postgres
mcptoon install github              # search and install

Results carry a ✓ (registry-verified), a type tag (npm / pypi / hosted / remote) and a call count. Data comes from two upstreams, queried live and never stored:

Why nothing is bundled: a built-in list would need a release to update and would pull someone else's source into your supply chain. mcptoon stores only a pointer — the search writes one line into your own ~/.mcptoon/config.json; third-party source never lands inside mcptoon.

Then distribute to every agent:

mcptoon sync          # push the new tools to every detected agent
mcptoon manifest      # see every tool (name index, the cheapest view)

How to read a result: the index mixes official @modelcontextprotocol/* servers with packages individuals publish. A ✓ means the registry verified the entry — your cue to read the source before you hand it credentials.

The skills half: skills search <query> queries the open skills index (skills.sh) — find a skill by what it does, then install it with skills add <git-url>. mcptoon installs the repos you point it at — the catalog is the open ecosystem itself.

Starter packs: if you'd rather not pick tool-by-tool, four built-in packs — essentials, web-research, code-review, docs — each bundle a few tools plus a ready-made prompt. mcptoon install --packs lists them; mcptoon install --pack essentials installs one. The two research packs need no API key.

The three bills, and why you must not mix them

mcptoon saves tokens in three separate places. Comparing the numbers across them is meaningless.

There are two ways to read "how much does this save?", and they answer different questions. The gateway figure — full schemas versus what the agent actually loads under the default compact exposure — is what installing mcptoon buys, and it is the honest headline (89% on the machine this was written on). The slim-schema figure (same tools, terser descriptions) is the smaller, secondary claim (88.5%), and it is what manifest --slim buys. mcptoon status prints both, side by side, from one measurement, so the two can never drift apart.

Bill 1 · Tool discovery (manifest): 99.2% smaller by default

mcptoon manifest with no flags is this tier. Want more? --slim (names plus param types, 8,282 tokens, −88.5%) or --full (the complete schema).

Bill 2 · Call results (call): optional, --toon saves ~6%

This bill comes due after a tool returns. mcptoon call prints JSON by default and saves nothing by default. Add --toon to shrink the result.

Bill 3 · Skill catalog (skills): 926,232 → 39 resident + 501 per lookup

mcptoon skills manifest                      # 39 tokens, resident
mcptoon skills resolve "make a PDF" --k 5    # 501 tokens, the 5 most relevant

One table, all three, on the machine this README was written on:

The fixed headline benchmark — a synthetic sample of 255 tools across 50 servers, tiktoken cl100k_base. It is not your catalog; run mcptoon status to see your own numbers:

Measure your own catalog with mcptoon bench (it reports your tools, not this fixed sample). The exact three-format table above is reproducible too — no clone needed:

mcptoon manifest                 # populate the schema cache
python -m mcptoon.bench_tokens   # the same JSON / --slim / --compact counts, your tools

Method and caliber: docs/tiktoken-benchmarks.md.

Install

pip install mcptoon

Linux (Debian/Ubuntu 23.04+, and any distro that follows PEP 668), and Homebrew Python on macOS. A plain pip install into the system Python is refused there with error: externally-managed-environment. That is the distro protecting itself, not a problem with mcptoon. Install it the clean way instead:

bash

pipx install mcptoon     # isolates the CLI; recommended
# or, into a virtualenv you control:
python3 -m venv ~/.mcptoon-venv && ~/.mcptoon-venv/bin/pip install mcptoon

pip install --break-system-packages mcptoon also works, but it writes into the system Python — prefer pipx or a venv. On Windows, pip install mcptoon just works.

# Run without installing (needs Node)
uvx mcptoon demo --quick

# Isolated CLI install (avoids the PEP 668 error on Linux / Homebrew Python)
pipx install mcptoon

# From source (for development)
git clone https://github.com/activeing123/mcptoon.git
cd mcptoon
pip install -e . --no-build-isolation

# Claude Code plugin
/plugin marketplace add activeing123/mcptoon

Works with every AI agent

mcptoon is a CLI tool — a manager, not a resident proxy or service, and not a client library. Your agent never talks to your upstream MCP servers directly: it either runs mcptoon commands, or reaches them through mcptoon's own MCP entry (mcptoon serve). Both are registered by quickstart; you don't pick one (see the table for which host takes which). So the config is written once and shared:

# Agent needs GitHub access mid-task? It just runs:
mcptoon add github --url https://api.githubcopilot.com/mcp/
# Done. No JSON editing. No restart. No lost context.

mcptoon serve is the other direction: run all your configured servers behind one MCP endpoint, with connection pooling and per-agent keys, for clients that insist on a proxy. You don't have to choose between the two — the CLI stays the primary way in, and the bridge is wired alongside it wherever the host can take one.

Why a CLI, not a proxy

MCP's premise is that every capability is a server your agent must be configured to reach — which is why one new tool means editing per-agent JSON in a different format for each, restarting everything, and re-paying the full schema cost in every agent.

A command line is the one interface every agent already has. And the form factor is measurably cheaper on its own, before mcptoon does anything:

• Firecrawl: the same task cost 1,365 tokens via CLI vs 44,026 via MCP — 32×
• Scalekit: CLI 10–32× cheaper, 100% reliable vs MCP's 72%

The proxy form exists too — mcptoon serve puts all configured servers behind one MCP endpoint — and quickstart wires it alongside the CLI, so you never have to choose. This section is about the shape mcptoon is built around, not a switch you have to flip.

All commands

mcptoon quickstart              # one-shot start (discover + configure + register the gateway)
mcptoon discover                # scan this machine for MCP servers (--write to keep, --health to probe)
mcptoon import                  # import servers from Claude Desktop / Cursor / Cline / Windsurf
mcptoon init                    # create a sample config (--auto to discover and fill it)
mcptoon list                    # show configured servers
mcptoon manifest                # all tool names (compact by default; 255 tools = 581 tokens)
mcptoon manifest --slim         # names + param types (8,282 vs 71,929 = −88.5%)
mcptoon inspect <server> <tool> # inspect one tool's schema
mcptoon search <query>          # search tools across servers
mcptoon call <server> <tool> '{"args":"here"}'   # call a tool
mcptoon add <name> --stdio|--http <cmd|url>     # add any MCP server
mcptoon remove <name>           # remove a server
mcptoon install <name> --npm|--pip|--url <pkg>  # install + auto-generate handler
mcptoon install --search <kw>   # search the live registries (nothing installed)
mcptoon update                  # refresh each server's cached tool surface; report what moved
mcptoon plugin install <dir>    # install an Agent Plugins 1.0.0 plugin
mcptoon sync                    # sync native config to every detected agent
mcptoon health                  # health-check every MCP server
mcptoon serve                   # run as an MCP server (stdio/HTTP)
mcptoon skills list             # list the skill catalog (--usage adds hit counts)
mcptoon skills sync <src>       # distribute a skill catalog to every agent's folder
mcptoon skills resolve "<task>" # BM25 shortlist of skills (offline, no LLM)
mcptoon bench                   # prove the savings on this machine (tools + skills, one table)
mcptoon demo                    # one command, live demo on your machine
mcptoon demo-server             # the same proof with zero downloads (11 stdlib tools)
mcptoon doctor                  # self-check: Python, config, connectivity
mcptoon status                  # one screen: what's configured, gateway wired, tokens saved
mcptoon stats                   # token-savings dashboard (vs raw JSON)
mcptoon report                  # the whole savings account: tools + skills + cumulative
mcptoon usage                   # local call statistics
mcptoon footer-facts            # one line of savings for a chat footer (never blocks)
mcptoon retrieve <handle>       # get back the original behind any --smart handle
mcptoon config                  # show gateway settings (footer, welcome, lang)
mcptoon toggle <server> <tool>  # enable/disable a single tool (--list to show all)
mcptoon policy                  # per-tool compression policy (raw / toon / slim)
mcptoon completion ps           # shell completion (bash/zsh/fish/powershell)
mcptoon off                     # remove the gateway entry from your agents (reversible)
mcptoon restore                 # drop the gateway and put your servers back
mcptoon uninstall               # full cleanup — prints the plan first (--dry to preview)

Full reference in DEVELOPERS.md.

Format family: four tiers, compact by default

All optional; the default is already the leanest tier.

Why compact by default? Choosing which tool to use only needs names (581 tokens for 255 tools); parameter detail matters at call time and is fetched on demand. Defaulting to full schemas would hand the 99.2% right back.

The rule for agents: use manifest to choose, inspect before you call. Measured on 41 live tools: an agent guessing arguments from names alone lands ~10% valid calls, while one that runs inspect once for the 2–3 tools a turn actually uses hits 100% — identical to injecting every schema, at a fraction of the cost.

Trust and safety

It's just a 341KB native CLI — delete it anytime; keep it, and you never have to configure tools or skills for any agent again. mcptoon touches your agent configs, so it is built to be transparent — and easy to walk away from.

Three guards run on every tool result before your agent sees it:

• Reversible. mcptoon restore undoes a takeover (drops the gateway and returns your servers); mcptoon off removes only the gateway entry; mcptoon uninstall --dry prints the full removal plan first. Your servers are never deleted unless you ask.
• No telemetry. No analytics, no crash reports, nothing phoned home.
• Local-first. Your tools, skills and files stay on your machine. The only thing that leaves is install --search, which asks the registries for a catalog listing — it never sends your data.
• No stored credentials. API keys pass straight from your config or environment.
• No dependencies. Pure Python standard library — nothing in the supply chain to audit. CI enforces it with scripts/check_zero_deps.py.
• No daemon. Pure CLI — no resident process, no listening port.
• Formats don't break compatibility. The wire protocol is always standard JSON-RPC; compact/slim/toon only affect mcptoon's output to the agent. If --toon decoding ever fails it falls back to JSON, and one --full restores the native schema. No lock-in.

Scope, in one line: mcptoon is an index and a config manager — it points you to each tool's own source, which you can review on its own terms.

Credits and references

• TOON standard — v4.1 (MIT), vendored from python-toon and credited in NOTICE
• ToonDeck — a GUI console for mcptoon (pre-alpha): same engine, point and click instead of typing commands

Who builds this: mcptoon is an independent third-party project maintained by @activeing123. It is not affiliated with Anthropic.

Contributing

git clone https://github.com/activeing123/mcptoon.git
cd mcptoon
pip install -e . --no-build-isolation
pip install pytest pytest-cov
python -m pytest tests/ -v   # 1901 passed, 2 skipped

Three hard rules: zero dependencies (CI-enforced), new behavior ships with tests, Windows is a first-class target. New here? Start with CONTRIBUTING.md and DEVELOPERS.md.

The codebase: 25,227 lines of Python across 43 modules, zero third-party dependencies.

License

Apache 2.0. See LICENSE and NOTICE.