Mesh Memory

Mesh Memory

Mesh Memory is a self-hosted semantic memory service with a built-in MCP server. It stores documents (worklogs, decisions, notes, research) in PostgreSQL with pgvector and retrieves them by meaning — a query like "what database did we pick?" surfaces a saved note that says "chose Redis for caching" even with zero keyword overlap. Embeddings are generated locally with multilingual-e5-base (768 dimensions); the core flow requires no external API keys.

When to Use

• Saving a session worklog, decision, or research note so a later session can find it.
• Recalling past work by topic when you do not remember the exact words you used.
• Sharing a long-lived knowledge base across multiple agents, terminals, or teammates.
• Organizing context by role or project through workspaces (one workspace per role/project).
• Looking up structured tags (e.g. all type:decision entries from one project).

Trigger keywords: remember this, recall, what did we decide, save worklog, persistent memory, cross-session, knowledge base, mesh_add, mesh_search, mesh_focus.

Prerequisites

1. A running Mesh Memory instance reachable from the MCP server. Local Docker is the common path:

   docker compose up -d
   

   See https://github.com/dklymentiev/mesh-memory for the full Quick Start.
2. The MCP server (mcp_server.py) registered with your client (Claude Code, Cursor, Claude Desktop, or any other MCP-aware agent).
3. MESH_API_URL pointing at the running instance (default: http://localhost:8000).
4. Python 3 available on PATH for the MCP server process.

Procedure

Step 1 — Register the MCP server

Add the server to your client's MCP configuration:

{
  "mcpServers": {
    "mesh": {
      "command": "python3",
      "args": ["/path/to/mesh-memory/mcp_server.py"],
      "env": {
        "MESH_API_URL": "http://localhost:8000"
      }
    }
  }
}

On Windows (PowerShell), adjust the path to the Windows-style location, e.g. C:\\path\\to\\mesh-memory\\mcp_server.py, and ensure python3 resolves (or use python).

Step 2 — Verify the instance is reachable

curl $MESH_API_URL/health

Expected output:

{"status":"healthy"}

If this fails, the MCP server cannot reach the backend — see Pitfalls.

Step 3 — Use the 13 MCP tools

When the server is reachable, these tools become available:

Step 4 — Save a session worklog

After completing work, persist it for future sessions:

mesh_add(
  content="Investigated 502s on the checkout flow. Root cause: missing CORS header on the cart API. Fix shipped in commit abc123.",
  tags="type:worklog,topic:checkout,date:2026-05-23",
  workspace="developer"
)

date: and source: are added automatically when omitted. Type and topic tags are inferred from nearest neighbors after the embedding completes — 5–10 seed documents are required before inference kicks in.

Step 5 — Recall past work by meaning

Search across sessions for related context, even with different vocabulary:

mesh_search(query="checkout was failing for some users", limit=5, workspace="developer")

The query shares no keywords with the original note ("502s", "CORS"), but embedding-based search surfaces it.

Step 6 — Switch role / context

For a multi-role agent, switch the active workspace at the start of a session:

mesh_focus(workspace="sysadmin", prefetch=true, limit=5)

Subsequent calls default to that workspace. Pin a role-prompt document at the top of each workspace so the agent re-orients on every prefetch.

Step 7 — Cross-workspace search with weights

To pull context from related domains without diluting the primary signal:

mesh_search(
  query="nginx rate limit recipe",
  workspaces={"sysadmin": 0.7, "security": 0.2, "developer": 0.1},
  limit=10
)

Results are merged across workspaces and re-scored by workspace weight.

Step 8 — Structured lookups by tag

When you need an exact filter rather than semantic similarity:

mesh_bytag(tags="type:decision,status:active,guid:my-project", limit=20)

Tag Conventions

Mesh accepts arbitrary tags. The recommended prefixes (used by auto-inference and surfaced by mesh_schema):

HARD RULE: With fewer than ~5–10 documents in a workspace, neighbor inference is skipped. Manually tag seed documents until the corpus self-organizes.

Examples

Example 1 — Save a decision

mesh_add(
  content="Chose Redis for caching layer. Evaluated Memcached but Redis gave us persistence and pub/sub for free.",
  tags="type:decision,topic:caching,guid:checkout-redesign",
  workspace="developer"
)

Example 2 — Recall that decision by meaning

mesh_search(query="what database did we pick for caching", workspace="developer", limit=3)

No keyword overlap with "Redis" or "Memcached," but semantic search surfaces the decision.

Example 3 — List all active decisions for a project

mesh_bytag(tags="type:decision,status:active,guid:checkout-redesign", limit=20)

Pitfalls

1. Tool calls fail with connection errors. The MCP server cannot reach MESH_API_URL. Verify the instance is up (curl $MESH_API_URL/health returns {"status":"healthy"}) and the env var is set in the MCP config block.

2. A saved document does not appear in semantic search yet. Embedding generation runs in the background. After a save, expect a 1–2 second delay before semantic search hits the new document. Use mesh_get(guid=...) to confirm the document exists immediately.

3. Search returns results from the wrong domain. The active workspace is not what you expected. Call mesh_focus(workspace="<name>") explicitly, or pass workspace= on every call. With no focus and no explicit param, calls land in the default workspace.

4. Auto-tagging never adds anything. The workspace has too few documents for neighbor inference (~5–10 minimum). Manually tag a handful of seed documents, then auto-inference takes over.

5. A deleted document still appears in a search result. Embedding indices are eventually consistent; rerun the search after a few seconds, or use mesh_get(guid=...) to confirm deletion.

6. Mesh is a knowledge store, not a chat memory. Long conversation transcripts should be summarized before being saved — do not dump raw transcripts.

7. High-precision lookups need tags, not vectors. Vector similarity is robust but not perfect; for exact structured queries, prefer mesh_bytag over mesh_search.

8. CPU-only embeddings by default. Very large corpora (hundreds of thousands of documents) benefit from a dedicated instance and pgvector tuning — not covered here.

9. Optional AI categorizer is disabled by default. It requires an OpenAI-compatible LLM endpoint. The core flow (local embeddings, search, tagging) works without it.

Verification

1. Check instance health:

   curl $MESH_API_URL/health
   

   Expected: {"status":"healthy"}

2. Confirm a saved document exists immediately after add:

   mesh_get(guid="<returned-guid>")
   

   Expected: the document content and tags are returned.

3. Confirm semantic search finds it (after 1–2 s):

   mesh_search(query="<paraphrased query with no keyword overlap>", limit=5)
   

   Expected: the document appears in results.

4. Check workspace stats:

   mesh_stats()
   

   Expected: document count, tag count, and workspace name match expectations.

5. Verify tag schema is loaded:

   mesh_schema()
   

   Expected: recognized prefixes (type:, status:, date:, source:, guid:) are listed.

Related Skills

• mcp-server-setup — general MCP server registration patterns across clients.
• pgvector-tuning — scaling pgvector for large embedding corpora (referenced when Mesh grows beyond CPU defaults).