carrel
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| carrel_searchA | Full-text search the carrel desk index (.carrel/carrel.db) under a root directory. Requires a prior |
| carrel_packB | Pack a file or directory into LLM-ready context: file tree plus extracted text of supported files (text, pdf, office, ebook). |
| carrel_inspectA | Metadata for one file: detected type, size, mtime, sha256, mime guess and per-type detail (pages, dimensions, headings, columns, ...). |
| carrel_tagB | Manage desk tags: add/rm/ls tags on a file, or find files by tags. |
| carrel_noteB | Attach a free-text note to a file in the desk db, or list its notes. |
| carrel_indexB | Build or refresh the desk full-text index under root (default: the whole root). Returns indexed/skipped/pruned counts. |
| carrel_convertA | Convert a file to another supported type. Text targets also return the converted content inline (capped at 1 MiB). |
| carrel_diffB | Compare two files (text / struct / pdf / image). |
| carrel_redactA | Redact patterns from a text file's contents and return the result. Never writes; PDFs must go through the CLI ( |
| carrel_doctorB | Environment report: external tools found (with versions or install hints), per-command status, and the capability table gating each command. |
| carrel_metaB | Typed key/value fields on desk files: set, get, ls or rm fields on a file, or find files by conditions (vendor=acme, total>1000, due<2026-11-01, paid?). |
| carrel_fieldsB | Extract vendor, invoice number, PO, dates, subtotal/tax/total, currency, IBAN and account last-4 from a document (invoice, receipt, statement) with a confidence per field; optionally save them as desk meta fields. |
| carrel_mailA | Email files (eml/mbox): save a message's attachments into a directory, or group the messages of files/directories into threads. |
| carrel_refsB | Find reference numbers (invoice, PO, order, check, account, tracking, ticket, IBAN, routing, EIN, VAT, ISBN, GTIN, DOI, UPS, USPS) in a file or directory; optionally tag files with ref:: or group files by shared value. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 14 tools
Most tools have clearly distinct targets: index/search/pack/inspect/convert/diff/redact/doctor are easy to tell apart by resource and action. The main overlap is between carrel_fields and carrel_refs, which both extract invoice/PO/reference data from documents (and both can tag or persist results), and to a lesser degree among carrel_tag, carrel_meta, and carrel_note, which all attach or query metadata on desk files.
All tools share a uniform carrel_ prefix and a single lowercase snake_case token, which is highly predictable. However, the second token is sometimes a verb (search, pack, inspect, convert) and sometimes a noun (meta, fields, mail, refs), and several tools multiplex multiple actions (tag add/rm/ls, meta set/get/ls/rm), so it isn't a strict verb_noun convention.
14 tools is squarely in the well-scoped range and each tool covers a distinct document-desk capability (indexing, retrieval, conversion, comparison, extraction, metadata, mail). Nothing appears redundant or padded for the stated scope.
The surface covers the lifecycle well: indexing, search, packing, inspection, conversion, diffing, redaction, tags/notes/meta, and structured extraction. Minor gaps exist — no way to delete a note or remove a tag-adjacent record, no direct file/content deletion or write-back for redaction (PDF redaction is CLI-only), and no plain browse/read primitive independent of pack.