What is provenance tracking in agent-memory-mcp?

Provenance tracking records the origin, timestamp, and context of each stored memory, enabling the server to weight recall by reliability and freshness. This helps agents distinguish between high-confidence facts and uncertain or outdated information.

How do I configure where memories are stored?

Set the MEMORY_DB_PATH environment variable to point to your SQLite database file. By default, it stores memories at ~/.agent-memory/memory.db. You can customize this path when installing or running the MCP server.

Does agent-memory-mcp support multi-agent scenarios?

The server uses a single SQLite database (configurable via MEMORY_DB_PATH), so multiple agents can share the same memory store. Provenance tracking helps distinguish which agent or context created each memory.

What is decay-weighted recall?

Decay-weighted recall automatically reduces the relevance score of older memories over time, so recent information is prioritized during retrieval. This mimics human memory and prevents stale data from dominating agent decisions.

Can I use agent-memory-mcp without a database setup?

No, the server requires a SQLite database. However, setup is minimal—just set MEMORY_DB_PATH to a writable location, and the server will initialize the database on first run.

MCP ServerSTDIOOfficialv0.1.3

Kira Autonoma Agent Memory MCP Server

Store and retrieve agent memories with automatic decay weighting and provenance tracking. Built for developers building multi-turn AI agents that need persistent, context-aware recall across conversations.

io.github.kira-autonoma/agent-memory-mcp

Try MCP with AI models View source

Hosted URL

Local install

Transport

STDIO

Auth

No auth required

What the Kira Autonoma Agent Memory MCP server does

How models use it and what it is built for.

Connect to Kira Autonoma Agent Memory

Local install — runs as a subprocess.

npx @kiraautonoma/agent-memory-mcp@0.1.3

Environment variables

Configuration this server reads at startup.

MEMORY_DB_PATH
Path to SQLite database file (default: ~/.agent-memory/memory.db)

Resources

Where to find authoritative docs and source for Kira Autonoma Agent Memory.

Source repository

Example prompts for Kira Autonoma Agent Memory

Paste any of these into Agent Studio after connecting Kira Autonoma Agent Memory.

Store this conversation context with provenance metadata for later recall
Retrieve memories from the last 7 days weighted by relevance decay
Show me all memories tagged with user_id:12345 and their confidence scores
What does the agent-memory-mcp README say about the feedback loop mechanism?

Kira Autonoma Agent Memory MCP server — FAQ

Common questions about connecting and running Kira Autonoma Agent Memory.

What is provenance tracking in agent-memory-mcp?
Provenance tracking records the origin, timestamp, and context of each stored memory, enabling the server to weight recall by reliability and freshness. This helps agents distinguish between high-confidence facts and uncertain or outdated information.
How do I configure where memories are stored?
Set the MEMORY_DB_PATH environment variable to point to your SQLite database file. By default, it stores memories at ~/.agent-memory/memory.db. You can customize this path when installing or running the MCP server.
Does agent-memory-mcp support multi-agent scenarios?
The server uses a single SQLite database (configurable via MEMORY_DB_PATH), so multiple agents can share the same memory store. Provenance tracking helps distinguish which agent or context created each memory.
What is decay-weighted recall?
Decay-weighted recall automatically reduces the relevance score of older memories over time, so recent information is prioritized during retrieval. This mimics human memory and prevents stale data from dominating agent decisions.
Can I use agent-memory-mcp without a database setup?
No, the server requires a SQLite database. However, setup is minimal—just set MEMORY_DB_PATH to a writable location, and the server will initialize the database on first run.

Skip the local setup — run MCP in your browser

MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 60+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.

Open Agent Studio