pi-cognee

Cognee AI memory for Pi — your coding agent with persistent, queryable knowledge graph memory.

Two backends, one interface:

Mode	How	Best for
SDK (default)	In-process via @cognee/cognee-ts	Zero-infrastructure, works out of the box
MCP	Remote Cognee MCP server	Shared memory across agents, self-hosted

Install

pi install npm:@kerryhatcher/pi-cognee

Quick Start (SDK mode)

SDK mode works immediately — no server needed. Just set your LLM API key:

# In pi:
/cognee-config llmApiKey sk-...
/cognee-config llmModel gpt-4o-mini

Then start using memory:

> Remember this: the project uses Postgres with Prisma ORM
> What database does this project use?

MCP Mode

If you run a Cognee MCP server, switch to MCP mode:

/cognee-mode mcp
/cognee-config mcpUrl http://localhost:8001/mcp

Commands

Command	Description
/cognee-mode [sdk|mcp]	Switch backend mode
/cognee-config	Show all config
/cognee-config <key>	Show one config value
/cognee-config <key> <value>	Set a config value

Tools

All tools work identically in both modes:

Tool	Description
cognee_health	Check connectivity
cognee_remember	Store text in memory
cognee_recall	Search memory
cognee_forget	Delete datasets or all memory
cognee_datasets	List datasets
cognee_dataset_data	List items in a dataset
cognee_create_dataset	Create a new dataset
cognee_client_info	Show client identity and mode
cognee_cognify_file	Ingest a file (base64)

Config Keys

Key	Default	Description
mode	sdk	Backend mode: sdk or mcp
mcpUrl	http://localhost:8001/mcp	MCP server URL
llmModel	—	LLM model (SDK mode)
llmApiKey	—	LLM API key (SDK mode)
embeddingProvider	—	Embedding provider (SDK mode)
embeddingModel	—	Embedding model (SDK mode)
vectorDbProvider	—	Vector DB provider (SDK mode)
graphDbProvider	—	Graph DB provider (SDK mode)

License

MIT