# XF Model Source

> Find the right AI model for your job. Free discovery and ranking; tests on your own key.

- name: dev.xpansion/xfms
- version: 0.5.3
- connection class: R0 (autonomous)
- trust: 96/100
- quality: 75/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/dev.xpansion%2Fxfms
- tools: https://api.protogrid.dev/v1/servers/dev.xpansion%2Fxfms/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "xfms": {
      "type": "http",
      "url": "https://xfms.xpansion.dev/mcp",
      "headers": {
        "X-OpenRouter-Key": "${X_OPENROUTER_KEY}"
      }
    }
  }
}
```

Placeholders to fill: `${X_OPENROUTER_KEY}`.

## Trust

- hygiene: 100
- liveness: 100
- freshness: 100
- provenance: 85
- drivers: +repository +dns-namespace +website +reachable +uptime-100% +updated-3d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 75/100 (good)

### protocol: 75 (weight 25)

- warn `protocol.modern`: newest supported version is 2025-11-25; 2026-07-28 not supported, older versions are deprecated until 2027-07-28
- n/a `protocol.stateless`: only applies to 2026-07-28 servers
- pass `protocol.transport`: streamable HTTP
- n/a `protocol.list_ttl`: only applies to 2026-07-28 servers

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 100 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 7 of 7 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 2,348 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 25 (weight 15)

- warn `stability.changes`: 11 tool changes in 30 days
- fail `stability.rug_pull`: 5 tools kept their name but changed most of their description in 30 days; review before trusting them

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 7, about 2,348 tokens to load them all
- tool set hash: 47b8c803cb1afaa83240fd7925cfee19
- tools last changed: 2026-10-16T21:47:06.833Z
- badge: [![protogrid quality](https://protogrid.dev/badge/dev.xpansion/xfms.svg)](https://protogrid.dev/servers/dev.xpansion/xfms)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-16 `watch_job` changed (description, inputSchema)
  - before: Saves a finished job for XF Model Watch, which re-tests it as new models arrive and emails the person only when a model passes every must-have and ranks higher on their priorities. Needs a signed-in subscriber. Ask the person first: Watch keeps the job and its example tasks (encrypted, deletable any time). Pass consent=true only if they agree. Pass the discover session, and either test_session from a finished comparison or the tasks, plus the model they use now.
  - after: Saves a finished job for XF Model Watch, which re-tests it as new models arrive and emails the person only when a model passes every must-have and ranks higher on their priorities. Needs a signed-in subscriber. Ask the person first: Model Watch keeps the job and its example tasks (encrypted, deletable any time). Pass consent=true only if they agree. Pass the discover session, and either test_session from a finished comparison or the tasks, plus the model they use now.
- 2026-10-16 `compare` changed (inputSchema)
- 2026-10-16 `watch_job` added
  - Saves a finished job for XF Model Watch, which re-tests it as new models arrive and emails the person only when a model passes every must-have and ranks higher on their priorities. Needs a signed-in subscriber. Ask the person first: Watch keeps the job and its example tasks (encrypted, deletable any time). Pass consent=true only if they agree. Pass the discover session, and either test_session from a finished comparison or the tasks, plus the model they use now.
- 2026-10-16 `benchmark` changed (inputSchema)
- 2026-10-15 `field_results` added
  - Results from real comparisons on Model Source for one kind of job (call with no job_type for the list): which models passed, how often, which won, typical cost per task and the most common failure. Counted without anyone's job details. A starting point; test on the person's own work before choosing.
- 2026-10-15 `discover` changed (inputSchema)
- 2026-10-15 `compare` changed (title, description, inputSchema, outputSchema) **meaning changed**
  - before: Run a live A/B test between 2–5 user-specified models for a stated purpose. NO ranking step — the supplied model_ids ARE the candidate set. Generates 5 representative test queries from the purpose, runs them through every named model in parallel, and returns real cost, latency, and plain-English commentary on who won what. Unknown IDs are dropped with a note; if fewer than 2 IDs resolve, the call refuses. Use this whenever the user names specific models to compare (e.g. 'A/B test X and Y'). For engine-chosen candidates, use `benchmark` instead. Costs more than `rank` (10+ live LLM calls). Free-tier note: when any candidate ends in ':free', the probe is capped at 3 queries (no adaptive expansion) because free-tier rate limits often push longer probes past the deploy's 5-minute ceiling — evidence will be shallower. The commentary surfaces this when it happens.
  - after: Same live test as benchmark, but on the models the person names (model_ids). Needs a finished discover session so every requirement is judged. Estimate first, then go=true.
- 2026-10-15 `benchmark` changed (title, description, inputSchema, outputSchema) **meaning changed**
  - before: Run a live A/B test against the engine's TOP 3 PICKS for a stated purpose — the engine chooses the candidates from the full catalog. Generates 5 representative test queries (auto-expands to 10 or 15 if results are too close to call), runs them through the picked models in parallel, and returns real cost, latency, and plain-English commentary on who won what. Use AFTER `pick` or `rank` when the user wants the engine's own picks stress-tested with live data. DO NOT use this when the user has already named specific candidate models — the engine will ignore the names and test its own picks. Use `compare` instead in that case. Costs more than `rank` (15+ live LLM calls).
  - after: The live test: the 5 proven models and 1 promising newcomer from discovery do the person's real tasks side by side, every requirement is judged, and the results pick the winner. First call (session + tasks) returns the estimate and an offer — nothing is spent. Show the person the cost; only when they agree, call again with offer and go=true. Runs on the person's OpenRouter key (X-OpenRouter-Key header). Save the returned report as a local file.
- 2026-10-15 `discover` changed (title, description, inputSchema, outputSchema) **meaning changed**
  - before: Show which quality dimensions matter for a stated purpose, WITHOUT ranking any models. Returns the inferred weights and the discovery-walk trace. Useful for understanding how XFMS interprets the purpose before committing to a pick.
  - after: Discovery: finds the true requirements of the person's job before any model is chosen. Start with purpose: the reply carries the discovery instructions and the evidence catalogue. YOU ask the person the questions (one thing each, plain words, offering the returned choices); then call again with the sealed session, their answers and your structured updates. Each reply gives progress, rejected updates with reasons, next targets, answer choices from live data and a shortlist. At most 5 rounds. Ends with a confirmation view; call with confirm=true once the person agrees. The person can stop any time (end_now=true). Nothing about the person is stored by XFMS.
- 2026-10-15 `pick` changed (title, description, inputSchema, outputSchema) **meaning changed**
  - before: Return the single best LLM for a stated purpose. Concise output, no list. Use when the user has settled on the criteria and just wants one answer.
  - after: The single best proven model for a job worked out with discover, with why and the runners-up. Pass the discover session.

Full history: https://api.protogrid.dev/v1/servers/dev.xpansion%2Fxfms/changes

## Remotes

- https://xfms.xpansion.dev/mcp (streamable-http, auth none, reachable true, uptime30d 1)

## Packages

_none_

## Tools (7)

- `benchmark`: The live test: the 5 proven models and 1 promising newcomer from discovery do the person's real tasks side by side, every requirement is judged, and the results pick the winner. First call (session + tasks) returns the estimate and an offer — nothing is spent. Show the person the cost; only when they agree, call again with offer and go=true. Runs on the person's OpenRouter key (X-OpenRouter-Key header). Save the returned report as a local file.
- `compare`: Same live test as benchmark, but on the models the person names (model_ids). Needs a finished discover session so every requirement is judged. Estimate first, then go=true.
- `discover`: Discovery: finds the true requirements of the person's job before any model is chosen. Start with purpose: the reply carries the discovery instructions and the evidence catalogue. YOU ask the person the questions (one thing each, plain words, offering the returned choices); then call again with the sealed session, their answers and your structured updates. Each reply gives progress, rejected updates with reasons, next targets, answer choices from live data and a shortlist. At most 5 rounds. Ends with a confirmation view; call with confirm=true once the person agrees. The person can stop any time (end_now=true). Nothing about the person is stored by XFMS.
- `field_results`: Results from real comparisons on Model Source for one kind of job (call with no job_type for the list): which models passed, how often, which won, typical cost per task and the most common failure. Counted without anyone's job details. A starting point; test on the person's own work before choosing.
- `pick`: The single best proven model for a job worked out with discover, with why and the runners-up. Pass the discover session.
- `rank`: The full shortlist for a job worked out with discover: 5 proven models and 1 promising newcomer, each with its standing on every requirement, plus near misses and what would change the order. Pass the discover session.
- `watch_job`: Saves a finished job for XF Model Watch, which re-tests it as new models arrive and emails the person only when a model passes every must-have and ranks higher on their priorities. Needs a signed-in subscriber. Ask the person first: Model Watch keeps the job and its example tasks (encrypted, deletable any time). Pass consent=true only if they agree. Pass the discover session, and either test_session from a finished comparison or the tasks, plus the model they use now.