Skip to content

MCP Tools

In normal use, you should not need to choose low-level tools by hand. The agent starts with agent_workspace or guide_next_steps, while LINZA keeps the working route: check the folder, index, show a map, build review cards, learn from accepted examples, and propose the next step in preview.

Default Tool Set

By default, LINZA exposes a compact MCP tool set of 15 tools.

ToolPurpose
agent_workspaceOne entry point for diagnostics, import, map, review, teaching, growth, relations, memory search, and context export
guide_next_stepsShow the next safe step in plain language
index_allIndex the Markdown folder into .linza/linza.db
searchSemantic and lexical search
read_fileRead an exact file from the working folder
get_statsQuick local database counters
scan_vaultRead-only folder diagnostic
build_review_apply_queueBuild review cards with stable rq-* IDs
approve_review_queue_itemsPreview or apply selected cards
list_approved_itemsShow accepted local records
explain_nodeExplain one node: links, bridges, context
explain_relationshipExplain a possible relation between two nodes
who_dependsShow dependencies and neighbors
show_flowFind a route or flow between nodes
create_context_packBuild a compact context package for an agent

agent_workspace

This is the main agent entry point. It contains working actions such as doctor, map, review, teach, grow, connect, and calibr. You should see a short status, evidence, next steps, and a clear explanation of what could change; raw JSON stays in the technical layer.

Review And Apply

LINZA uses stable rq-* IDs. That makes application exact: only cards you have seen should be accepted. If IDs do not match, stop and rebuild the preview.

Advanced Tool Set

For development and debugging, advanced tools can be enabled:

powershell
$env:LINZA_TOOL_SURFACE="advanced"

Normal work does not need this. Keep the path short: diagnostics, index, map, review cards, teaching, and the next preview.

{Semiotronika}
Telegram · Volnaya Sreda · Email