Midas runs on your own machine — the client starts it, so there is no endpoint to ping. 78 installs a week from pypi. Last commit 28 Jul 2026.
Local-first, source-traceable agent memory — no LLM at ingest, fully offline
We read the source, 3 d ago · rules 995398eba69f
Found in continuous integration, deployment or infrastructure files, or in a neighbouring package of the same monorepo. None of this is installed when you add the server: it describes how the project is built and released. We list it because a leaked key in a build pipeline is still a real problem, but it is not something this server does on your machine.
subprocess.run(remove, capture_output=True, text=True) # best-effort; ignore if not present
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
Today is the operative word: we check Midas every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add midas -- uvx midas-memory-mcp
{
"mcpServers": {
"midas": {
"command": "uvx",
"args": [
"midas-memory-mcp"
]
}
}
}
[mcp_servers.midas]
command = "uvx"
args = ["midas-memory-mcp"]
{
"mcpServers": {
"midas": {
"command": "uvx",
"args": [
"midas-memory-mcp"
]
}
}
}
{
"servers": {
"midas": {
"type": "stdio",
"command": "uvx",
"args": [
"midas-memory-mcp"
]
}
}
}
Optional settings from the registry entry, left out of the block above:
MIDAS_MCP_DB (Path to a SQLite file to persist memory across restarts (default: in-memory)), MIDAS_MCP_EMBEDDER (Embedding backend: 'local' (bge ONNX, offline) or 'hashing' (default: local)), MIDAS_MCP_MAX_RECORDS (Cap the store; above it the lowest-value memories are auto-forgotten (no LLM)), MIDAS_MCP_MIN_IMPORTANCE (Relevance floor 1-5 for auto-capture; turns scoring below it are skipped).
Trace AI agent execution: every tool call, every error, every dollar. Open source, local-first.
Detect stale source context and retain traceable code-chain understanding for AI coding agents.
Open-source agent that observes and fixes your application. Query logs, traces, metrics, incidents.
Bounded working memory for coding agents — typed nodes, current-vs-stale state, MCP-mountable.
Trace an image to a clean SVG and export a full app-icon set, locally, from your coding agent.
Stock research memory for Claude: theses, sourced claims and SEC filing alerts in a local graph
Local-first PHI detection for source code, LLM prompts, logs, and analytics
LLM Orchestration Traced Agent
Answers built from our own checks of this server.