llm-engram

Durable, explainable long-term memory for Simon Willison's llm CLI, powered by Engram (Lumetra).

llm-engram registers six tools that let any model running under llm store and recall facts across sessions, with semantic + knowledge-graph retrieval handled server-side.

Install

llm install llm-engram

(or, from a checkout: llm install -e .)

Configure

Get an API key at https://lumetra.io, then export it:

export LLM_ENGRAM_API_KEY=eng_live_...
# Optional: pin a default bucket so you don't have to repeat it
export LLM_ENGRAM_DEFAULT_BUCKET=my-project
# Optional: self-hosted Engram
export LLM_ENGRAM_API_BASE=https://engram.internal.example.com

ENGRAM_API_KEY is also accepted for parity with other Engram clients; LLM_ENGRAM_API_KEY wins if both are set.

Tools

Usage

List the tools:

llm tools

One-shot:

llm -T engram_store_memory "Remember: I prefer pytest over unittest."
llm -T engram_query_memory "What testing framework do I prefer?"

Chat with memory wired in:

llm chat -T engram_store_memory -T engram_query_memory \
         -T engram_list_memories -T engram_list_buckets

Per-call bucket override:

llm -T engram_store_memory \
    "Q2 OKR: ship v2 of the ingest pipeline" \
    --tools-arg engram_store_memory bucket=work

Self-hosting

Point LLM_ENGRAM_API_BASE at any Engram-compatible REST endpoint. The plugin uses only the documented /v1/... routes.

Privacy

See PRIVACY.md.

License

MIT — see LICENSE.