v2.0.0 live on npm

GraphRAG memory.

npm install. Done.

Your AI forgets everything. We fix that.

$npm install -g engramport

Not on MCP? Import api.engramport.com/openapi.json into ChatGPT Custom GPT Actions, Gemini, n8n, Zapier, anything that speaks HTTP.

For any AI agent, any tool. BYO LLM. Cryptographic provenance.

Get your API key →View on GitHub

Free tier: 1,000 engrams + 1 namespace. No credit card.

Two install paths, one memory

However your AI tools talk, EngramPort listens.

For MCP clients

$ npm install -g engramport

Claude Desktop, Claude Code, Cursor, Continue, Cline, Zed, Goose, and any MCP-compatible client. One install, three env vars, working memory.

For everything else

→ import api.engramport.com/openapi.json

ChatGPT Custom GPT Actions, Gemini Extensions, n8n, Zapier, LangChain. Anything that imports an OpenAPI spec. Same namespace, same provenance seal.

One Eidetic substrate underneath both. Same memory, same audit trail, same BYO LLM. The transport is just delivery.

Works with everything

MCP-native via npm. OpenAPI-importable for everything else. Same memory across Claude, ChatGPT, Cursor, Gemini, n8n, and anything that speaks HTTP.

Bring your own LLM

Anthropic, OpenAI, Google, DeepSeek, Mistral. Your provider key, your inference bill. We never touch your model choice.

Cryptographic provenance

Every read and write is signed. Built on AEGIS — the same audit substrate Covenant Systems uses for regulated AI deployments.

Not just similarity. Structure.

Semantic search returns direct matches plus graph-connected context. Find what you searched for, plus what it links to.

Multi-agent shared memory

Multiple agents on the same namespace see each other's writes. Cross-session coordination for AI swarms — the use case single-vendor memory products can't serve.

Open registry, private data

npm install is fully public; your engrams live in tenant-isolated namespaces with row-level security. The substrate is yours, even when the install path is shared.