Total Agent Memory
Local persistent memory for coding agents, with a temporal knowledge graph
- Pricing
- Open source
- Platforms
- CLI, self-hosted, macOS
- Runs
- Locally
- Version
- 15.0.0
- Listing updated
- 2 days ago
About Total Agent Memory
Total Agent Memory is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
Coding agents forget decisions between sessions, so conventions get explained again. This server keeps decisions, fixes, facts, errors and session summaries in a SQLite database on your own machine and returns them to any MCP client. The site's example saves a decision to use pgvector over ChromaDB in one session, then recalls the reasoning days later in another repo.
Recall fuses full-text BM25, locally computed embeddings, fuzzy matching and a knowledge graph with reciprocal rank fusion, then reranks with a local cross-encoder. The default profile makes no LLM call on write or search. Facts can carry validity intervals, and a newer value for a single-valued fact retires the older one. The release notes mention 79 tools. The maker reports 0.607 recall at 5 on LoCoMo and 0.448 on BEAM at 1M tokens, measured on 13.x.
The same package can run as a team server with personal, team and shared areas, roles, authorship and an audit trail. Setup needs Python 3.11 or newer, and the first start downloads about 330 MB of models. A setup wizard registers it with clients such as Claude Code, Codex, Cursor, Windsurf and Gemini CLI. It is MIT-licensed, with no telemetry; paid plans buy support and installation.
Install
Run in a terminal
claude mcp add total-agent-memory -- uvx total-agent-memoryAdd to claude_desktop_config.json
{
"mcpServers": {
"total-agent-memory": {
"command": "uvx",
"args": [
"total-agent-memory"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"total-agent-memory": {
"command": "uvx",
"args": [
"total-agent-memory"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"total-agent-memory": {
"type": "stdio",
"command": "uvx",
"args": [
"total-agent-memory"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"total-agent-memory": {
"command": "uvx",
"args": [
"total-agent-memory"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"total-agent-memory": {
"command": "uvx",
"args": [
"total-agent-memory"
]
}
}
}Tools this server provides (4)
Taken from the project documentation; the running server may differ.
- memory_save
- Save a decision, fact or other memory
- memory_recall
- Recall memories matching a query
- kg_add_fact
- Add a fact to the knowledge graph with validity
- kg_at
- Query the knowledge graph as of a point in time
What it does
- Local-first storage
- Memories live in SQLite files on your machine, with embeddings computed locally.
- Hybrid recall
- Combines BM25, embeddings, fuzzy matching and a knowledge graph, then reranks.
- Temporal facts
- Facts carry validity intervals and newer values can retire older ones.
- No LLM in default path
- Saving and searching make no LLM call, so retrieval can be benchmarked offline.
- Team server
- Personal, team and shared memory with roles, authorship and audit history.
- Setup wizard
- Detects and registers with Claude Code, Codex, Cursor, Windsurf, Gemini CLI and others.
- Private text stripping
- Private blocks and credentials are stripped on every write path.
Good for
- Carry architecture decisions across sessions and repos
- Stop re-explaining project conventions to an agent
- Share team knowledge through a self-hosted server
- Run a local memory baseline for research
Pricing
Open source
MIT-licensed and free; paid support from $4,800 per year
Prices change. Check totalmemory.dev before you buy.
Total Agent Memory FAQ
Is it free?
The software is MIT-licensed and free; paid plans cover support, installation and response commitments.
Does it call an LLM?
Not in the default save and search path; optional features use a provider you configure, which can be local.
What does it need?
Python 3.11 or newer; the first start downloads about 330 MB of models.
Is hosting offered?
Not yet; it is built to run on your own infrastructure.
Is Total Agent Memory free?
Yes. Total Agent Memory is open source, so you can use and self-host it for free.
How do I add Total Agent Memory to Claude Code?
Run this command in your terminal: claude mcp add total-agent-memory -- uvx total-agent-memory
What are the best alternatives to Total Agent Memory?
The closest alternatives to Total Agent Memory are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.
Alternatives to Total Agent Memory

REA
MCP serverCLI and MCP server that lets coding agents reverse engineer apps and binaries
Runs locally- Open source
- 464K GitHub stars
- Developer tools

Codebase Memory
MCP serverLocal MCP server that builds a code knowledge graph for AI coding agents
Runs locally- Open source
- 48.7K GitHub stars
- Knowledge and memory

GitHub
MCP serverOfficial MCP server connecting AI assistants to GitHub repos, issues and workflows
Not checked yet- Remote
- Open source
- 34.5K GitHub stars
- Developer tools

Serena MCP: the IDE for your agent
MCP serverMCP toolkit giving coding agents symbol-level code retrieval and editing
Runs locally- Open source
- 31.5K GitHub stars
- Developer tools

XERJ
MCP serverLocal search engine for AI agents with cited answers over your code and docs
Runs locally- Open source
- 19.2K GitHub stars
- Knowledge and memory

Ouroboros
MCP serverSpec-first layer that pins requirements and verifies results of AI coding agent work
Runs locally- Open source
- 6.4K GitHub stars
- Developer tools