CognetivyCognetivy is an open-source state layer for AI coding assistants that adds workflows, run tracking, and persistent collections - all local.
Everything your agent needs to work well
Cognetivy gives your agent an operational layer: workflows, runs, events, and structured collections in a local workspace. No LLMs inside - just the data and tools your editor's agent uses via Skills and MCP.
Core features
Six building blocks that give your agent memory, structure, and visibility.
Studio
A read-only browser UI that opens with one command. Inspect workflows, runs, events, artifacts, and all collected data - locally.
- Workflow canvas (DAG view)
- Run + event timeline
- Collections explorer
Define workflows
Give your agent an explicit process to follow - step by step, every time. Workflows are versioned JSON files that live in your project.
- DAG-based node structure
- Versioned workflow.json
- Reusable across runs
Track runs and events
Every run captures a full timeline: what your agent decided, what tools it called, what it produced, and how long it took.
- Per-run event log
- Status and duration tracking
- Jump to any run instantly
Structured collections
Keep ideas, sources, and outputs in schema-backed collections that persist and accumulate across runs - downloadable as CSV.
- Schema-enforced data
- Accumulates across runs
- Download as CSV
Memory + process manager
The model is the brain, your editor is the workspace - Cognetivy is the persistent memory and process layer that ties them together.
- Persistent local workspace
- Structured .cognetivy/ folder
- Works across sessions
Skills & MCP
Works with any MCP-compatible client via the standard protocol. Agent skills provide richer, zero-config integration.
- Standard MCP protocol
- Agent skills included
- No lock-in, no proprietary format
Great for
Cognetivy works best when your agent needs to do the same thing more than once.
Repeatable workflows
Define once, run many times. Keep your agent's process consistent and auditable across every session.
Structured research
Run multi-step research with coding agents. Capture sources, summaries, and plans in one traceable run.
Traceability for teams
Every output links back to its source and reasoning. Essential for compliance, review, and collaboration.
Works with your tools
No proprietary format, no new editor. Cognetivy plugs into the agents you already use via Skills and the standard MCP protocol.
Claude Code
Full skills + MCP support
OpenClaw
Skills + MCP support
Cursor
MCP server integration
MCP & agent skills
Standard protocol, zero lock-in
Any client that supports MCP works out of the box. See setup docs →
Ready to try?
Get structured workflows and full observability for your AI agent in under a minute.