Getting Started

MCP Tools

The Promemo MCP tool surface for loading, searching, indexing, previewing, and saving memory.

MCP Tools

Promemo v0.4.0 expands the MCP tool surface so AI coding tools can use project memory directly.

promemo_load_context loads prompt-ready memory for a feature.

{
  "feature": "authentication",
  "related": true,
  "token_budget": 1200
}

promemo_search returns exact matching Markdown lines from saved memory.

{
  "query": "refresh token"
}

promemo_search_context runs retrieval search over chunked Promemo memory.

{
  "query": "why did we choose passkeys",
  "mode": "hybrid",
  "limit": 5
}

Supported retrieval modes:

keyword
semantic
hybrid

Use hybrid for most natural-language project questions.

Index And Project Map

promemo_index_status reports whether the generated local search index exists, plus index version, chunk count, feature count, file relationship count, and decision count.

promemo_index_rebuild rebuilds reproducible cache data under:

.promemo/cache

promemo_related_features returns features related to a target feature through shared files or decisions.

{
  "feature": "authentication"
}

promemo_project_map returns the local feature, file, decision, and related-feature graph.

Memory Writes

promemo_preview_memory previews a structured memory write without changing files.

promemo_save_memory saves reviewed structured memory.

promemo_extract_memory uses the configured provider to extract structured memory from unstructured text. It defaults to dry-run behavior.

{
  "feature": "authentication",
  "source_text": "We changed token refresh handling...",
  "dry_run": true
}

promemo_memory_schema returns the MemoryInput JSON schema.

Repository Helpers

Promemo also exposes:

  • promemo_list_features
  • promemo_tree
  • promemo_doctor
  • promemo_snapshot
  • promemo_read_feature_file

Allowed feature files include:

memory.md
notes.md
context.md
architecture.md
decisions.md
api.md
data-model.md
todos.md
prompts.md

Example AI Workflow

  1. User asks an assistant to work on authentication.
  2. Assistant calls promemo_load_context for authentication.
  3. Assistant calls promemo_related_features to discover nearby context.
  4. Assistant makes code changes.
  5. Assistant calls promemo_snapshot to inspect git state and TODOs.
  6. Assistant drafts memory.
  7. Assistant calls promemo_preview_memory.
  8. User reviews.
  9. Assistant calls promemo_save_memory.

A local-first memory layer for AI-assisted projects.

© 2026 Promemo
Local-first project memory for AI-assisted development.