KMP MCP Server
by underpass-aiio.github.underpass-ai/kmpv0.24.0
Local-first agent memory that preserves what happened, when and why
context tax
queued
security
queued
cold start
queued
freshness
Active2d ago
Install KMP MCP server
No published package or hosted endpoint yet — see the repository README for build-from-source instructions.
Configuration
| Variable | Required | Secret | Description |
|---|---|---|---|
| KMP_MCP_DATA_DIR | — | — | Optional memory directory; otherwise KMP uses the project or per-user default |
| KMP_VIEWER_ADDR | — | — | Optional viewer address, or off to disable the local viewer |
| KMP_KERNEL_GRPC_ENDPOINT | — | — | Optional remote KMP endpoint; omit it to keep the kernel in this process |
Freshness
Active — last maintenance signal 2d ago. The newest of the signals below sets the band.
Last commit (default branch)
2026-10-07 · 2d ago · GitHub
Latest release
2026-10-04 · 5d ago · GitHub · v0.25.0
Package published
no data · npm/PyPI
Registry entry updated
2026-09-27 · 12d ago · official registry · v0.24.0
FAQ
›Does KMP require an API key?
No required environment variables are declared in its published metadata.
›Can I use KMP as a remote (hosted) MCP server?
No hosted endpoint is published; it runs locally over stdio.
›Is KMP in the official MCP registry?
Yes, as io.github.underpass-ai/kmp.
Alternatives to KMP
Other memory MCP servers.
MemoryActiveKnowledge graph-based persistent memory for maintaining context.
HindsightActiveHindsight: long-term memory for AI agents modeled on human memory.
Codebase MemoryActiveIndexes a repo into a persistent local knowledge graph of functions, classes, call chains, HTTP routes and cross-service links, using tree-sitter and LSP.
CogneeActiveMemory manager for AI apps and agents using graph and vector stores, with ingestion from many data sources.
HonchoActiveMemory that reasons: continual learning for stateful agents. Better context, fewer tokens.
Basic MemoryActiveLocal-first AI memory: a semantic knowledge graph of Markdown files that humans and LLMs both read and write, with optional cloud sync and team workspaces.