Projects and Tables

Tool	Use this to
`tamr_list_projects`	List the projects you can access, with the most recently updated first. Each project includes its name, display name, whether it is the default project, and operational metadata.
`tamr_list_tables`	Page through System of Record tables, oldest first. `project_id` (from `tamr_list_projects`) is optional. You can omit it to list the first page of tables in every project. Optionally, use `page_size` (default 10, max 100, per project), `cursor` (requires project_id), and `data_product_id` to filter to one data product.
`tamr_get_table_schema`	Return a table's record structure, including attribute names and data types, as a JSON Schema document. You can supply either the `table_id` or `table_name`.
`tamr_describe_table`	Return a table's searchable fields (with human-readable hints), search configurations, and match labels. Call this before `tamr_search` to learn valid field names.

Records

Tool	Use this to
`tamr_search`	Perform fuzzy entity-resolution search. Supply the `table_id` or `table_name` and search criteria (field → value pairs; partial values match). This tool supports pagination, searching across all of a table's configurations at once (`search_config ="all"`), and optionally attaching source records or history. On an unknown field, it returns the valid field list, enabling a client to self-correct.
`tamr_find_records`	Perform an exact-match lookup. Supply the `table_id` or `table_name` and match (field → exact value; all fields must match). Matching is case, whitespace, and type-sensitive, with no fuzzy matching. For list fields, pass a list: `{"country": ["US"]}`. Results are paginated (default 10, max 100).
`tamr_get_records`	Fetch up to 200 records by ID in one call (`golden rec_`… IDs or external IDs), fetched in parallel. This tool tolerates partial failure. Results are bucketed into found / not-found / errors.
`tamr_record_history`	Return a record's paginated change history, with newest events first.
`tamr_record_relationships`	Return a record's incoming and outgoing relationships across tables, with a human-readable label per relationship.

Note on search confidence:
tamr_search results are scored by entity-resolution match confidence (HIGH / MEDIUM / LOW), and indicates how likely a result is to be the same real-world entity as your query. Search is not a keyword-relevance score. A broad or single-attribute query (for example, just a state) will legitimately score LOW on every result, because the records cannot be confidently identified as "the same entity" from sparse information. Add more identifying attributes (name plus email, address, and so on) to raise match confidence.

Curation Queues

Tool	Use this to
`tamr_list_queues`	Find curation queues and their associated queueIds. By default, this tool returns active queues. You can filter by `project_id` (omit for all projects), `status`, `queue_type`, or `table_id`. Set `include_open_items` to count each queue's open items (exact up to 500; beyond that, reported as "500+").
`tamr_suggest_edit`	Suggest a correction to a record. This does not change the record directly. The suggested change is queued for a data steward to approve or dismiss. Requirement: The table must have an existing `SUGGESTED_RECORD_EDIT` queue.
`tamr_suggest_duplicate`	Suggest merging two records in the same table, with the candidate record merging in to the primary record. This does not merge the records. The suggested change is queued for a data steward to approve or dismiss. If the pair already has an open item, in either direction, nothing is submitted and the existing item is returned. Requirement: The table must have an existing `SUGGESTED_DUPLICATE_RECORD` queue.
`tamr_list_queue_items`	List items in a curation queue `OPEN`, `CLOSED`, or `ALL`).
`tamr_close_queue_item`	Apply or dismiss a suggested edit or duplicate item. `APPLY` commits the change to the master record(s) (applies an edit, or merges the duplicate). `DISMISS` discards the suggestion.

Pipeline Jobs

Tool Use this to

Tool	Use this to
`tamr_list_jobs`	List data pipeline jobs, including data loads, data product refreshes (`UPDATE`), and publishes, with most recent jobs first. You can filter by `job_type`, `state`, `errored`, `data_product_id`, `workflow_id`, `source_id`, `destination_id`, and `created_after` / `created_before` (ISO 8601). Results are paginated (default 10, max 30); pass the same filters on every page.
`tamr_get_job`	Get the current status of one job, any errors, and the status-history timeline. Supply a `job_id` from `tamr_list_jobs`.

tamr_list_jobs

List data pipeline jobs, including data loads, data product refreshes (UPDATE), and publishes, with most recent jobs first.

You can filter by job_type, state, errored, data_product_id, workflow_id, source_id, destination_id, and created_after / created_before (ISO 8601).

Results are paginated (default 10, max 30); pass the same filters on every page.

tamr_get_job

Get the current status of one job, any errors, and the status-history timeline.

Supply a job_id from tamr_list_jobs.

Notes on jobs:

A job in DONE state with no errors indicates that the job succeeded.
status.error.reason explains why a job ended or was cancelled with an error: EXECUTION_FAILED (it ran and failed), PLANNING_FAILED (it failed before it started), or STOPPED_BY_USER (someone cancelled the job; not a failure).
Some update jobs are associated only with a workflow, not with a data product. A data_product_id filter does not match those jobs.

Prompts

Prompts are multi-step workflows exposed as slash commands in the client:

Prompt	Use this to
`investigate_entity`	Find an entity by free-text query, then pull its full record, history, and relationships.
`explore_relationships`	Map the network of relationships around a record, optionally filtered to one relationship type.
`suggest_correction`	Find the relevant record and matching curation queue, and walk through submitting a suggested edit.

Resources

Read-only resources mirror the read tools for clients that support the MCP resources primitive:

Resource URI	Returns
`tamr://table/{table_id}/schema`	The table's searchable fields, hints, configs, and match lables.
`tamr://table/{table_id}/metadata	The table's name, domain/data-product labels, and record count.
`tamr://record/{table_id}/{record_id}`	A full record, by golden ID or external ID.