# codelined-ag/Extracto

Source: https://glimind.com/server/github%2Fcodelined-ag%2Fextracto
Registry id: `github/codelined-ag/extracto`
Status: unknown

## Tools (66)

- `dropbox_import`: Download a file from the user's Dropbox and submit it for OCR. Returns a jobId. Supports pdf, png, jpg, webp up to 32 MiB.
- `dropbox_list_folder`: List a Dropbox folder. Pass an empty path or '/' to list the App folder root. Returns each entry's kind (file/folder), name, path, size, and modified timestamp.
- `dropbox_push`: Push a COMPLETED job's text back into Dropbox in the chosen format. The filename comes from the original job; pass `folder` to control where it lands ('/' or em
- `e2e_encrypt`: Encrypt a chunk of text with the user's registered RSA public key using AES-256-GCM + RSA-OAEP-SHA256 envelope encryption. Returns the sealed envelope. The user
- `e2e_status`: Get the user's current E2E key registration status: whether a public key is registered, its fingerprint, and when it was registered.
- `google_drive_import`: Download a file from the user's Google Drive (by file id) and submit it for OCR. Returns a jobId. Supports pdf, png, jpg, webp up to 32 MiB.
- `google_drive_list_folder`: List a Google Drive folder. Pass an empty string or 'root' to list the My Drive root. Returns each entry's id, name, mimeType, kind (file/folder), size, and mod
- `google_drive_push`: Push a COMPLETED job's text back into Google Drive in the chosen format. Pass `parentId` to control the destination folder (empty falls back to My Drive root).
- `integration_disconnect`: Disconnect a cloud provider (Dropbox / Google Drive / OneDrive). Best-effort revokes the OAuth token at the provider before deleting the local row.
- `integrations_list`: List the user's connected cloud-drive integrations and which providers the operator has configured at the instance level.
- `integrations_status`: Show which cloud integrations (Dropbox, Google Drive, OneDrive) are available on this server and which the user has connected. Returns availability flags + each
- `job_edit_page`: Replace the markdown text of a single page on a COMPLETED job. The previous text is appended to the per-page edit history; the job's extractedText is re-stitche
- `job_equations`: Extract LaTeX equations from a COMPLETED job. Display equations are detected as $$..$$, inline as $..$. Best results when the job ran with documentPreset='acade
- `job_export`
- `job_form_fields`: Extract structured form fields from a COMPLETED OCR job whose result includes a form-shaped block (typically when documentPreset='form' was used). Returns each 
- `job_page_history`: Fetch the edit history for one page on a job. Each entry has the prior text, characterCount, and editedAt timestamp; the newest entry is first.
- `job_set_tags`: Replace the set of tags applied to a job. Pass an empty array to clear all tags.
- `job_stop`: Request a running job to stop and pause at the next page checkpoint.
- `jobs_bulk_tag`: Apply tags to many jobs at once. Default mode 'add' is a union (idempotent on duplicates). Mode 'replace' clears existing tags on each job and writes the new se
- `jobs_list`: List the caller's recent OCR jobs (newest first). All filters are AND-combined; tagIds is OR-combined within itself.
- `kb_export`: Chunk + embed + push a completed OCR job's text to a vector store. Strategies: fixed (char-window with overlap), sentence (sentence-merge), paragraph (paragraph
- `kb_search`: Full-text search across the caller's KB-exported jobs.
- `kb_test_connection`: Probe any supported vector store (Chroma, Qdrant, Weaviate, Milvus, OpenSearch, Pinecone, Typesense) for reachability and auth before running a KB export. Retur
- `keys_create`: Create a new API key. Returns the plaintext key once; store it now.
- `keys_delete`: Revoke an API key by id.
- `keys_list`: List the caller's API keys (id, label, scopes, createdAt, lastUsedAt). Secrets are never returned.
- `keys_rotate`: Rotate an API key in place. Returns the new plaintext key once; the previous secret is revoked. The id stays stable so dependent callers can update via Authoriz
- `keys_update`: Update an existing API key's name, scopes, or rateLimitPerMinute without revoking it. Pass any subset of fields to change. Pass rateLimitPerMinute=null to clear
- `metrics_get`: Aggregate usage and queue metrics for the caller (job counts, processing time, OCR pages, by-day buckets, by-provider/model).
- `oauth_app_clear`: Delete the user's saved OAuth credentials for a provider. Falls back to server-wide creds if any.
- `oauth_app_set`: Save the user's own OAuth client_id + client_secret for a provider. Stored encrypted. Overrides any server-wide credentials.
- `oauth_app_status`: Show whether the user has personal OAuth credentials saved for a provider, falling back to server creds. Returns source ('user' | 'server' | 'none'), clientIdLa
- `ocr_compare`: Submit one input file to N models in parallel and get a comparisonId. Use ocr_comparison_get to fetch the per-model outputs and the server-computed word-level d
- `ocr_comparison_get`: Fetch a comparison by id. Returns each model's job + extractedText, plus all-pairs word-level diffs (N(N-1)/2 per comparison).
- `ocr_estimate`: Estimate the dollar cost of running OCR on a set of pages before submitting. Pricing is fetched live (OpenRouter, LiteLLM mirror), static (Mistral OCR per-page 
- `ocr_get`: Fetch a single OCR job (id, status, extractedText, structured result).
- `ocr_submit`: Submit one or more files for OCR. Returns the batch id and per-file job ids; poll ocr_get to retrieve results.
- `onedrive_import`: Download a file from the user's OneDrive (by item id) and submit it for OCR. Returns a jobId. Supports pdf, png, jpg, webp up to 32 MiB.
- `onedrive_list_folder`: List a OneDrive folder. Pass an empty string to list the App folder root. Returns each entry's id, name, kind (file/folder), mimeType, size, and modified timest
- `onedrive_push`: Push a COMPLETED job's text back into OneDrive in the chosen format. Pass `parentId` to control the destination folder (empty falls back to the App folder root)
- `pii_redact`: Run server-side PII redaction over a chunk of text. Replaces emails, phones (>=7 digits), Luhn-valid credit cards, IBAN-shaped strings, IPv4 addresses, URLs, da
- `presets_create`: Create an output preset (named bundle of provider/model/settings/postProcessing).
- `presets_delete`: Delete an output preset by id.
- `presets_list`: List the caller's output presets.
- `recommendations`: Get model recommendations per document type based on the user's recent OCR history. For each document type seen, returns the highest success-rate model + altern
- `s3_export`: Upload a completed OCR job's markdown + JSON to the user's pre-configured S3 bucket. Credentials live on the Extracto server (Settings → S3) and never round-tri
- `s3_list`: List objects in the user's configured S3 bucket under a sub-prefix. By default, only OCR-able file extensions (.pdf, .png, .jpg, .jpeg, .webp, .tif, .tiff, .bmp
- `saved_searches_delete`: Delete a saved search by id.
- `saved_searches_list`: List the caller's saved History searches with their filter payloads.
- `saved_searches_rename`: Rename a saved search without rewriting its filters. 409 if another saved search already uses the new name.
- `saved_searches_save`: Create or update a saved search. Idempotent on `name`: re-using an existing name overwrites that search's filters.
- `tags_create`: Create a tag, or update its color if a tag with the same name already exists.
- `tags_delete`: Delete a tag. The tag is also removed from any jobs it was applied to.
- `tags_list`: List the caller's tags with the count of jobs each tag is applied to.
- `tags_update`: Rename a tag and/or change its color. At least one of `name` or `color` is required.
- `watchers_create`: Create a watched folder. Extracto sweeps it on the chosen interval and submits any new file (pdf, png, jpg, webp; up to 64 MiB) to the OCR queue using the chose
- `watchers_delete`: Delete a cloud watched folder. Already-ingested file fingerprints are removed too; if the same file appears again it will be re-ingested.
- `watchers_list`: List the user's cloud watched folders (Dropbox / Google Drive / OneDrive). Each entry surfaces provider, name, folderPath, model, intervalSeconds, active, lastP
- `watchers_update`: Update one watched-folder. Pass only the fields to change. Re-activating clears any auto-pause failure counter.
- `webhooks_create`: Register a new webhook. Returns a one-time HMAC signing secret in the response. Save it now; it is not shown again. The receiver must verify the X-Extracto-Sign
- `webhooks_delete`: Delete a webhook. Cancels all pending retry deliveries.
- `webhooks_deliveries`: List recent delivery attempts for a webhook. Each row has the event, statusCode, ok, attempt, status (pending/delivered/exhausted), and durationMs. Useful for d
- `webhooks_list`: List the user's registered webhooks. Each entry includes id, url, events, active, lastFiredAt, failureCount.
- `webhooks_rotate_secret`: Rotate the HMAC signing secret for a webhook. Returns the new plaintext secret once; the previous secret is immediately invalidated. Update the receiver before 
- `webhooks_test`: Fire a synthetic signed delivery to a webhook so you can verify URL, signature, and firewall before relying on a real job.
- `webhooks_update`: Update a webhook. Pass any of active, url, or events to change them. To rotate the signing secret, use webhooks_rotate_secret.

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.
