Skip to main content

Reference

Tools your AI agent can use.

A connected agent can ask Kin about code and its recorded connections. This reference lists the available tools and their inputs. Set up your client first, then use the list to inspect a particular tool.

How this works in Kin

These are the tools the Kin MCP server serves, read from the server itself rather than written down. An agent that has been through setup can call every one of them against your repository.

Where this came from

Check the source version and tool profile

Generated on 2026-10-13 from kin-mcp v0.8.6, by starting the server the way a client starts it and asking it to list its tools. The profile is agent-default, which is what kin setup configures, and it serves 11 tools. Current release: v0.8.6.

Which tool to reach for

Start with locate to find something by name or by what it does, context for a bundle around an entity or a question, and refs for callers, importers and references. The tools are named the way the kin CLI names the same capability. Every response carries a freshness and coverage envelope, so read that before acting on an absence: a tool that found nothing and a graph that has not finished answer in the same shape.

The tools

  • context

    Around X? Code/routes by name/question.

    View inputs for context

    arguments: depth, details, entities, entity_id, focal_body, max_output_bytes, neighbor_bodies, question, token_budget

  • history

    Who changed X, when, approved?

    View inputs for history

    arguments: entity_id, limit, offset | required: entity_id

  • impact

    X changes? Callers/overrides first.

    View inputs for impact

    arguments: base, change_ids, cursor, dependents, entity_ids, head, include_tests, max_output_bytes, scope

  • locate

    Find matches

    View inputs for locate

    arguments: cursor, details, ignore_case, include_tests, kind, language, limit, max_output_bytes, mode, pattern, query, scope, sort

  • mutate

    Commit entity changes atomically, in one call.

    View inputs for mutate

    arguments: operations, summary, transaction_id | required: operations

  • neighbors

    X's deps/dependents? Paths `trace`.

    View inputs for neighbors

    arguments: depth, direction, entity_id, limit, max_output_bytes | required: entity_id

  • refs

    Uses X? References/overrides. Paths: `trace`.

    View inputs for refs

    arguments: cursor, details, entity_id, include_tests, max_output_bytes, proof_for_caller, query, relation_kinds, scope, view

  • source

    X's code? Exact, by entity_id.

    View inputs for source

    arguments: entity_id, full_body | required: entity_id

  • status

    Graph health; not needed to trust a reply. last_settled_selected_graph is older.

    View inputs for status

    arguments: cursor, max_output_bytes

  • trace

    X's reach? Direct+indirect callers in 1 call. to: route; scope: module.

    View inputs for trace

    arguments: cursor, depth, details, direction, focal, include_body, include_tests, limit_per_step, max_output_bytes, scope, to

  • transaction

    Plumbing for writes; `mutate` with transaction_id stages.

    View inputs for transaction

    arguments: action, message, scope, transaction_id

Next action

Call one of these from the client you already use.

The setup page configures every client Kin detects and proves each entry answers before it says it is done.

Connect your existing agent

Related