Claude Code

Memory for Claude Code

Persistent graph memory for Claude Code. MCP-native. One CLI command. Coding memory that persists across projects and restarts.

The problem

Claude Code is powerful inside a session, but the session ends. Architecture decisions, debugging context, PR history - all gone on the next run. Graphory gives Claude Code a persistent graph to query and write to from the terminal, so your coding agent actually remembers what you built and why.

Why Graphory with Claude Code

MCP-native for the terminal

Claude Code speaks MCP over streamable HTTP. Graphory plugs in with one claude mcp add-json command - no wrappers, no glue code.

Memory that outlives sessions

Architecture decisions, PR reviews, and debugging context persist across every new Claude Code run, every restart, every project.

Deterministic

No vectors, no embeddings, no hallucinated recall. Typed nodes, typed edges, exact traversals.

Temporal

Every memory carries a timestamp, a source, and an audit trail. Claude Code can answer when a decision was made and why.

Shared with every other client

The same graph is visible to Claude Desktop, Cursor, Windsurf, and any other MCP client. Terminal memory is not siloed.

Two-minute install

Get an API key, run one CLI command, restart your session.

Step 1 - Generate an API key

Sign in at graphory.io, go to Settings > API Keys, and create a key. It starts with gs_ak_.

Step 2 - Register Graphory with Claude Code

claude mcp add-json graphory '{
  "type": "http",
  "url": "https://api.graphory.io/mcp",
  "headers": {
    "Authorization": "Bearer gs_ak_your_key"
  }
}'

Graphory tools become available in your next Claude Code session. Anthropic's CLI flags evolve - check the integration page for the current shape.

What you can do

Real queries Claude Code can run once Graphory is connected.

"Remember that we picked Postgres over DynamoDB for the billing service because of transactional reads."

"What did I decide about the rate-limit strategy last month? Pull the thread."

"Who on the team has worked on the auth service, and what did they change?"

Measured, not marketed.

Deterministic extraction and retrieval, benchmarked against public datasets.

0.9107 accuracy

LongMemEval (agent + MCP)

0.8667 accuracy

LoCoMo-MC10 (agent + MCP)

0.900 F1

BizLineItemBench

Using Claude Desktop too?

Graphory works the same way in Claude Desktop - the same graph, with a drop-in MCP config.

Memory for Claude Desktop

Claude Code, with memory.

Free up to 100K nodes. No credit card.