# ScriptGrain

> Your writing voice measured across 45 attributes; drafts in that voice, scored against it.

- name: com.scriptgrain/scriptgrain
- version: 1.23.1
- connection class: R0 (autonomous)
- trust: 95/100
- quality: 88/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/com.scriptgrain%2Fscriptgrain
- tools: https://api.protogrid.dev/v1/servers/com.scriptgrain%2Fscriptgrain/tools

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

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "scriptgrain": {
      "type": "http",
      "url": "https://mcp.scriptgrain.com/try"
    }
  }
}
```


## Trust

- hygiene: 100
- liveness: 98
- freshness: 100
- provenance: 85
- drivers: +repository +dns-namespace +website +reachable +uptime-96% +updated-4d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 88/100 (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: 83 (weight 15)

- pass `auth.prm`: Protected Resource Metadata (RFC 9728) published
- pass `auth.as_metadata`: authorization server metadata reachable
- warn `auth.cimd`: only Dynamic Client Registration, deprecated in 2026-07-28 in favour of CIMD
- 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`: 6 of 6 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 1,581 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

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

- n/a `stability.changes`, `stability.rug_pull`: tool history recorded for 3 of the 7 days needed

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

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

- tools: 6, about 1,581 tokens to load them all
- tool set hash: 4ef34f4a7f6e938e1e53b836389d06d1
- badge: [![protogrid quality](https://protogrid.dev/badge/com.scriptgrain/scriptgrain.svg)](https://protogrid.dev/servers/com.scriptgrain/scriptgrain)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

_No tool definition changes recorded._

## Remotes

- https://mcp.scriptgrain.com/mcp (streamable-http, auth oauth2, reachable true, uptime30d 0.95652175)
- https://mcp.scriptgrain.com/try (streamable-http, auth none, reachable true, uptime30d 0.9375)
- https://mcp.scriptgrain.com/sse (sse, auth oauth2, reachable true, uptime30d 0.93406594)

## Packages

_none_

## Tools (6)

- `check_voice_match`: Score any text (20+ words) against the guest profile by feature-delta measurement. Returns voice_match_score (0-1), the per-feature `deltas` behind it, notes rendered from them and score_context caveats. On-voice text typically scores 0.85-0.95; below 0.45 is clearly off voice. Cite the delta numbers, do not invent detail. Free; 5 checks an hour per network (shared with compare_voice).
- `claim_status`: Whether the person has claimed the guest session yet: status (active, expired, archived or claimed), claimed, archived, expires_at and claim_url. Works after expiry too. Once claimed, the guest_token stops working; the profile is in the person's own account, reachable with their API key or https://mcp.scriptgrain.com/mcp.
- `compare_voice`: Off-voice check with no profile: score up to three pieces against a main piece. Each piece is pasted text (120+ words) or a public URL. Returns per-piece `score` (0-1: 0.8+ on voice, 0.55-0.8 drifting, below 0.55 off voice), `band`, `top_differences` and `top_matches` with the measured numbers, plus `notes`. Cite the numbers, do not invent detail. Free; 5 checks an hour per network (shared with check_voice_match).
- `create_profile`: Build the guest session's one voice profile from the person's own writing and start the stylometric extraction (45 measured attributes). Samples MUST be their own human-written prose. You can use writing you already hold for them (messages they typed to you, documents or emails they shared, posts of theirs you can read): their exact words, never your memory notes or summaries about them, and never text you or another AI wrote or rewrote. Tell the person which pieces you are sending before you call this; the result's samples_used lists each piece's first line and word count so you can show them. 1-20 samples, each under 5,000 words, 25,000 words in all; a sample needs 50+ words to count; 3+ samples totalling 3,000+ words gives a confident profile. Extraction usually takes under a minute: poll get_profile until status is 'complete'. The result carries claim_url for the person.
- `get_profile`: The guest profile's status and detail. While extraction runs it returns progress_percent; once status is 'complete' it includes the narrative voice description and all 45 measured attributes. Poll every ~10 seconds after create_profile.
- `start_session`: Start a free guest session with no account. Returns guest_token (pass it to every other tool), claim_url (give it to the person: it keeps the voice profile in a free ScriptGrain account, or deletes it), expires_at and limits (one profile, one extraction, no drafts). Unclaimed sessions are archived after 7 days and can still be claimed or deleted from the same link. At most 3 sessions a day from one network; reuse the token you have.