First Project Setup
Once MemStrata is installed, these steps take it from idle to actively saving you tokens.
Step 1 — Index your codebase
Section titled “Step 1 — Index your codebase”Navigate to your project root and run:
cd ~/my-projectmemstrata indexMemStrata builds a local graph of your codebase — symbols, dependencies, call graphs, import chains. This is stored entirely in ~/memstrata/index.db.
Expected time:
- ~500 files → 10–20 seconds
- ~5,000 files → 60–90 seconds
- ~50,000 files → 5–10 minutes (subsequent startups are incremental, under 5 seconds)
Step 2 — Start the harness
Section titled “Step 2 — Start the harness”memstrata serveThe harness starts on localhost:8080. Leave this terminal open, or add memstrata serve to your shell profile / system startup (see install guide for your OS).
Verify it’s running:
curl http://localhost:8080/health# {"status":"ok","version":"5.2.x","uptime_seconds":4}Step 3 — Connect your AI tool
Section titled “Step 3 — Connect your AI tool”Pick your integration:
- Cursor / Windsurf / VS Code → MCP server setup (2 config lines)
- Claude Code → stdio MCP setup (1 config block)
- Aider → harness proxy (
--openai-api-baseflag) - Claude.ai / ChatGPT → browser extension
Step 4 — See your first savings
Section titled “Step 4 — See your first savings”Make a few AI requests through your tool. Then open the dashboard:
memstrata dashboardThe Money tab shows savings in real time. After your first few turns you’ll see:
- Tokens sent (actual) vs. baseline
- Dollar savings this session
- Retrieval recall and precision
It takes 5–10 turns for the metrics to stabilize. The first turn includes the cold-start graph query — subsequent turns in the same session are faster.
Step 5 — Set your provider rate (optional)
Section titled “Step 5 — Set your provider rate (optional)”MemStrata uses a default rate table for cost calculations. If you have a negotiated rate or use a local model, set it:
memstrata config set provider anthropicmemstrata config set provider-model claude-3-7-sonnet-20250219memstrata config set provider-rate 3.00 # per million input tokensThe dashboard will use your custom rate for all dollar calculations from that point on.
Next steps
Section titled “Next steps”- Telemetry & Savings — understand the four metrics in detail
- Trust & Security — verify exactly what leaves your machine (short answer: almost nothing)
- FAQ — common setup questions