# Record what the agent learned: descriptions, tags, glossary terms or card edits

> Record what the agent learned: descriptions, tags, glossary terms or card edits: 1 companies fully do it, 7 with a limit, 26 carry the tag. Every listing in Company knowledge & data catalogues with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/company-knowledge.catalogue-write-back/
- Markdown: https://www.anchorterminal.com/niches/company-knowledge.catalogue-write-back/index.md (~1,700 tokens)
- Slim: https://www.anchorterminal.com/niches/company-knowledge.catalogue-write-back/index.min.md (~330 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/company-knowledge.catalogue-write-back/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

One company's docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 1 | 7 | 26 | 6 | 3 |

## What counts as done

The write is applied or held for review as reported, a preview is shown before it, failures come back as errors, and the change is findable by search within the index lag.

How agents ask for it: add a description to this table; tag these columns as PII; update the onboarding card with the new process

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Hexis Cloud (Bevelites GmbH)**: Agents write on a draft branch that opens a change request people approve in the app, check canWrite first, and get a saved-locally error when sharing is refused, with an instruction not to report it as shared. Changes are greppable once on the branch.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Atlan (Atlan Pte. Ltd.)**: Write tools return a preview and wait for approval, and read-only mode removes them on request, but bad glossary input returned an error with isError false that agents took for an answer (acc-32).
- **Collibra (Collibra Inc.)**: create_asset and edit_asset write assets with destructive annotations, and data-quality writes have a confirm preview, but edit_asset has no preview, failures come back in an error field, and search lag after a write is not stated.
- **DataHub (Acryl Data, Inc. (d/b/a DataHub))**: Write tools are off unless TOOLS_IS_MUTATION_ENABLED is set, carry readOnlyHint false, and proposals hold changes for review, but save_document and update_description report success while writing nothing (issues 211, 200) and saved documents miss search (acc-26).
- **Dust (Permutation Labs)**: The API upserts documents into data sources (under 120 a minute) and MCP tools create files, so learned text can be stored and synced in minutes, but no preview, review hold or error contract for writes is documented.
- **Guru (Guru Technologies, Inc.)**: MCP tools Create Draft and Update Card, which suggests updates, keep edits for review. No preview step or error behaviour is documented, and user tokens can write directly through REST.
- **Marmot (Marmot Data Ltd)**: Writes preview the exact before and after with confirm false and refuse with isError true without the manage permission, but confirm is a plain flag the server doesn't tie to a review, and the tools carry no destructive annotations.
- **OpenMetadata (Collate, Inc.)**: create_entity, patch_entity and create_lineage write under the user's role, and operator tool restriction is merged (issues 36577, 37161), but patch_entity reports a write held for approval as applied (issue 37757) and there is no preview step.

## Couldn't check

The docs were closed to us, or silent. These carry no mark either way.

- **Cohere North (Cohere Inc.)**: Automations and connectors could write to apps, but no readable page describes writes with preview and errors; the docs are closed by robots.txt.
- **Precisely AI Studio (Precisely Incorporated)**: Whether any governance asset writes descriptions or tags isn't on the readable page; the help centre couldn't be read.
- **Secoda (Secoda, Inc.)**: The MCP server registers no write tool; the REST API's edit endpoints were described in an earlier reading but were not re-read under Secoda's Terms of Use.

- not this job: Alation (Alation, Inc.): Release 1.0.0 removed update_catalog_asset_metadata and check_job_status from the SDK, leaving no metadata write tool in the local server.
- not this job: Basedash MCP (BaseDash Inc. (Delaware, United States)): Writes create and edit dashboards and charts, not catalogue descriptions, tags or glossary terms.
- not this job: ClicData MCP (ClicData SAS): The MCP page says building in ClicData from the AI client is coming later, so nothing writes descriptions, tags or terms today.
- not this job: Gemini Enterprise (Google LLC): Not aimed at recording descriptions, tags or cards; no write-back with preview and errors is documented for the API in the pages read.
- not this job: Glean (Glean Technologies, Inc.): Writes are artifacts saved to a user's Library and connected-tool actions through run_tool, not descriptions, tags or cards on indexed content.
- not this job: Onyx (DanswerAI, Inc.): The MCP scope is read:search only, with no write tools; documents pushed through the Ingestion API were not visible to MCP search (acc-38).
- not this job: Overclock (Overclock Technologies Limited): The server is read-only: tools retrieve and never create, change or delete, under scope engine:read.
- not this job: Procore AI Go (Procore Technologies, Inc.): The support page says the app cannot create or update Procore items or take other actions.
- not this job: Lens (Thought Industries) (Thought Industries, Inc.): The API creates, edits and syncs sources, but nothing writes descriptions, tags or card edits into the knowledge itself.

## What stops agents finishing

- write tools listed with no read-only switch (OpenMetadata)
- a write held for approval reported as applied (OpenMetadata issue 37757, maintainer-filed)
- confirm flag the server does not tie to a review (Marmot)

## Around it

Before: company-knowledge.find-data-asset, company-knowledge.lineage-trace. After: company-knowledge.find-data-asset.

## Where this comes from

Weight 6 of 100 in Company knowledge & data catalogues: 3 of 38 counted accounts (DataHub documents not searchable after write, Atlan glossary errors reported as success, DataHub document search empty); vendor write tools in Atlan, Collibra, Secoda, OpenMetadata, Marmot, Hexis and Guru

As data: https://www.anchorterminal.com/api/v1/use-cases/company-knowledge.catalogue-write-back.json, or the MCP tool get_use_case with id company-knowledge.catalogue-write-back. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
