Tools Overview

Use the tools in layers: discover files first, then run exact or semantic search, then confirm with file reads. When memory is enabled, recall it before analysis and capture meaningful outcomes after the work is done.

Main tools

Discovery

• localnest_usage_guide
• localnest_server_status
• localnest_health
• localnest_list_roots
• localnest_list_projects
• localnest_project_tree

Indexing and updates

• localnest_update_status
• localnest_update_self
• localnest_index_status
• localnest_index_project
• localnest_summarize_project

Search

• localnest_search_files
• localnest_search_code
• localnest_search_hybrid
• localnest_get_symbol
• localnest_find_usages

Memory

• localnest_task_context
• localnest_memory_status
• localnest_memory_recall
• localnest_memory_list
• localnest_memory_get
• localnest_memory_store
• localnest_memory_update
• localnest_memory_delete
• localnest_capture_outcome
• localnest_memory_capture_event
• localnest_memory_events

Verification

• localnest_read_file

Choose by goal

Goal	Start with	Then use
Find feature/module location	localnest_search_files	localnest_project_tree, localnest_read_file
Find symbol definition/export	localnest_get_symbol	localnest_read_file
Find symbol call/import usage	localnest_find_usages	localnest_read_file
Match exact code/symbol	localnest_search_code	localnest_read_file
Retrieve by concept/intent	localnest_index_status, localnest_index_project	localnest_search_hybrid, localnest_read_file
Rehydrate prior context	localnest_task_context	localnest_memory_recall (if needed)
Capture outcome for future tasks	localnest_capture_outcome	localnest_memory_events

Response format

Tools support response_format: "json" or "markdown". Responses also include meta.schema_version so MCP clients can detect contract revisions explicitly.

Tool success vs meaningful evidence

A successful MCP call means the tool executed. It does not always mean the response contains useful evidence.

• localnest_search_code returning no matches is a successful call, but weak evidence for the current task.
• localnest_read_file returning line content is meaningful evidence.
• Retrieval misses now include a recommended_next_action so the client can pivot immediately instead of stopping at an empty result.

Typical workflow

1

Check runtime state

localnest_server_status, localnest_task_context, and localnest_update_status confirm active backend, memory state, relevant recall, and version state.

2

Scope the workspace

localnest_list_roots and localnest_list_projects tell you where to search.

3

Recall prior context

If memory is enabled, prefer localnest_task_context before indexing or deep analysis.

4

Prepare the index

localnest_index_status and localnest_index_project make hybrid retrieval useful.

5

Retrieve and verify

localnest_search_hybrid narrows candidates, then localnest_read_file confirms exact lines.

6

Capture durable outcomes

After a fix, decision, review, or preference discovery, emit localnest_capture_outcome when memory is enabled.