Commands

What each command does, with the output of a real run.

Install

Run it without installing

npx efaimo weigh --client claude-code

Pin it in a project

npm i -D efaimo

Needs Node 22 or later. No account and no config file. For another client, swap claude-code for cursor, vscode or claude-desktop. The source is on GitHub under Apache 2.0.

efaimo weigh

Counts the tokens a server or skill adds to every request.

It serializes tool definitions the way your host does, then counts them. Point it at a server command, a URL, a repo, a skill folder, or a whole client config with --client.

npx efaimo weigh "npx -y @modelcontextprotocol/server-everything" real run · 2026-11-16
efaimo v0.6.2
weigh mcp  npx -y @modelcontextprotocol/server-everything
tools 16   resources 8   prompts 4

context cost of tool definitions (o200k tokens, estimated)
  raw JSON           2,443
  Claude-style       1,778   (~0.2% of a 1M window)
  OpenAI tools       1,791
  server instructions  831

heaviest tools (Claude-style)
   1. gzip-file-as-resource       243   desc 72 | schema 160
   2. get-header-params           177   desc 57 | schema 109
   3. trigger-log-messages        130   desc 93 | schema 26
   4. trigger-resource-update     126   desc 57 | schema 59
   5. trigger-list-changed        118   desc 53 | schema 54
   6. get-resource-reference      105   desc 11 | schema 83
   7. get-annotated-message       104   desc 14 | schema 78
   8. get-env                     103   desc 66 | schema 26
      (+8 more)
      block framing                     85   (<functions> wrapper; per-tool lines + this = total)

note: token counts are estimates using the o200k_base tokenizer; hosts add fixed framing text on top (see docs/METHODOLOGY.md)

The token cost page is built from this run.

efaimo check --mcp

Grades an MCP server and lists what to change for the 2026-07-28 spec.

The grade covers description quality and annotations. Spec readiness prints under it as an ungraded migration diff, and every rule names the SEP it comes from.

npx efaimo check --mcp "npx -y @modelcontextprotocol/server-everything" real run · 2026-11-16
efaimo v0.6.2
check mcp  npx -y @modelcontextprotocol/server-everything
grade A (95)   quality: 0 errors  1 warning  0 info

  ! E122  tool "echo": description misses 3/4 quality axes (length 40..600; says when to use it; mentions the result)

2026-07-28 readiness  1 item to migrate (a migration diff, not graded)
  ! E104  server declares the logging capability; MCP Logging is deprecated in 2026-07-28 (SEP-2577) and logging/setLevel is removed
          fix: log to stderr (stdio) or use OpenTelemetry; per-request level arrives via _meta io.modelcontextprotocol/logLevel

rules: https://github.com/efaimo-ai/efaimo/blob/main/docs/RULES.md

The official MCP reference server, checked by the published release.

efaimo check --skill

Grades a skill: frontmatter, size, trigger collisions, broken references.

Collisions are checked against every skill you have installed. It also flags prompt injection patterns.

npx efaimo check --skill ./SKILL.md real run · 2026-11-16
efaimo v0.6.2
check skill  efaimo
grade A (100)   0 errors  0 warnings  0 info

  no findings. clean.

rules: https://github.com/efaimo-ai/efaimo/blob/main/docs/RULES.md

efaimo grading its own SKILL.md. A release does not ship unless this passes.

efaimo test Experimental

Runs a task with and without a skill and reports the difference.

A lint pass cannot tell you whether a skill helps. This can, at the cost of real model calls on your own API key.

Two skills, run with and without live runs 2026-07-17 · claude-sonnet-5 · 8 trials per arm

"csv-cleanup helps on a messy CSV"

+0points
with skill 8/8 without 8/8
examples/scenario.example.yaml

"contoso-crm-import helps on an unknowable format"

+100points
with skill 8/8 without 0/8
examples/scenario.crm.yaml

Both skills lint clean. The first made no difference: the model passed without it. The second encodes a file format the model cannot guess, and only a run with and without it shows that.

Quoted from the repo's live runs. Re-running them costs API money; the scenarios are committed, so anyone with a key can.

efaimo mcp

Lets your agent check and weigh skills itself, over MCP.

A read-only MCP server with two tools, efaimo_check_skill and efaimo_weigh_skill. It reads files and spends nothing.

node dist/cli.js check --mcp "node dist/cli.js mcp" real run · 2026-11-16
efaimo v0.6.2
check mcp  node dist/cli.js mcp
grade A (100)   quality: 0 errors  0 warnings  0 info

  no quality findings. clean.

2026-07-28 readiness  clean, nothing to migrate

rules: https://github.com/efaimo-ai/efaimo/blob/main/docs/RULES.md

Run from the repo checkout. The published equivalent is npx efaimo check --mcp "npx -y efaimo mcp"

efaimo's own MCP server, audited by its own check --mcp.