Give your agents persistent memory, automatic context, project management, and observability.
No memory between sessions — agents start from zero every time
No project awareness — agents hallucinate file paths and structure
No observability — you don't know what they cost or where they fail
Vendor lock-in — switching providers means rewriting everything
No team knowledge — each developer's agent is an island
Agents remember across sessions
Identity + project facts injected at start
Claude, Gemini, Codex, OpenCode
Tokens, costs, tools, errors tracked
Federated brains share knowledge
Session hooks — inject & save
Universal CLI adapter & REST API
Memory
Projects
Observability
"If two products need to sync to function, the design is wrong."
Hot, warm, and cold tiers keep context relevant and compact.
Semantic search across all stored knowledge in milliseconds.
Works with Claude Code, Cursor, Windsurf, and any MCP client.
Personal and shared team brains. Recall auto-federates.
Browse entities, search memories, and manage your brain visually.
Persistent memory server with SQLite, FTS5, and a built-in React dashboard.
Tasks live in Markdown files. Human-readable, git-friendly.
Search, trace, and analyze your codebase through natural language.
FTS5-powered search over every project, task, and document.
Visual board with drag-and-drop, task metadata, and Markdown editing.
Every change is a clean diff. Review project state in any editor.
AI-native project manager. Markdown files are the source of truth; SQLite is just an index.
Input, output, and cache tokens broken down for every session.
Accurate pricing for every model across all supported providers.
Which tools get called, how often they fail, and where agents get stuck.
Ingest transcripts from Claude, Gemini, Codex, and OpenCode.
Visual cost trends, usage heatmaps, and per-project breakdowns.
Agent observability server with token tracking, cost estimation, and analytics dashboard.
Your brain's knowledge is available from the first prompt.
Conversations preserved before the context window shrinks.
One command: dbrain connect claude
Claude Code, Gemini CLI, Codex CLI, OpenCode.
Switch between Claude, Gemini, Codex, and OpenCode with a flag.
Memory + workspace + templates assembled automatically.
Full CLI parity via HTTP. Bearer token auth, configurable port.
Ask single-shot or chat in a persistent REPL with history.
Memory operations — recall, remember, entities, facts, search, and federation.
Project management — projects, tasks, docs, search, sync, and overview.
Model invocation — prompt execution, streaming, and multi-provider routing.
Observability — sessions, costs, tool analytics, and transcript ingestion.
Zero-config examples: cd examples && npm start
Install the core packages globally from npm.
Initialize and start your personal brain.
Connect your brain to your AI coding CLI.
Initialize and start the project manager.
Open your CLI. The full harness is active.
Each package does one thing well. Memory is memory. Projects are projects. No god objects.
Everything works from the terminal. Dashboards are a visual convenience, not a dependency.
Zero external dependencies. No Postgres, no Redis, no Docker. Just a single file on disk.
One server, two protocols. MCP for agents, REST for everything else. No port sprawl.
Data lives in files you can read, edit, and version. The database is just an index.
Four adapters, one interface. Switch providers with a flag. Your data stays with you.