Founder & CTO at Cognifyr
I build open-source infrastructure that helps AI agents remember context, retrieve evidence, interact with software, and execute actions under explicit human control.
| Layer | Project | What it provides |
|---|---|---|
| Context | Qarinah · GitHub · Paper · DOI | Evidence-linked project memory that turns large histories into compact, cited context for AI and coding agents. |
| Retrieval | Cockroach Crawler · GitHub · Docs | Bounded crawling, browser rendering, structured extraction, and provenance-preserving web evidence for agents and RAG systems. |
| Interaction | Cockroach Browser · GitHub · Paper · DOI | Local-first Chromium runtime for scoped sessions, semantic interaction, evidence capture, and controlled browser actions. |
| Governance | Maqam · GitHub · Docs | Policy checks, exact approvals, registered execution, and verifiable receipts for AI-agent actions. |
Together, these projects form an open-source infrastructure stack for agents that need durable context, real web evidence, controlled software interaction, explicit authority, and auditable execution.
The goal is not simply more autonomy. It is making agents useful in environments where control, provenance, security, and accountability matter.
AI Agent
│
▼
┌─────────────┐
│ Qarinah │
│ │
│ Context │
└──────┬──────┘
│
▼
┌─────────────────────────────┐
│ │
▼ ▼
┌───────────────┐ ┌───────────────┐
│ Cockroach │ │ Cockroach │
│ Crawler │ │ Browser │
│ │ │ │
│ Retrieval │ │ Interaction │
└───────┬───────┘ └───────┬───────┘
│ │
└──────────────┬──────────────┘
│
▼
┌─────────────┐
│ Maqam │
│ │
│ Governance │
│ & Execution │
└─────────────┘
Qarinah preserves the context an agent needs to continue work.
Cockroach Crawler retrieves bounded, source-linked evidence from the web.
Cockroach Browser gives agents controlled interaction with real browser environments.
Maqam governs higher-authority actions through policy, approvals, execution boundaries, and receipts.
Through Cognifyr, I have built enterprise knowledge assistants with SSO and permission-aware retrieval, private on-premises AI for healthcare research, real-time voice agents, workflow automation, and multi-tenant SaaS platforms.
Founder & CTO at Cognifyr · B.Tech in Computer Science, CUSAT · Winner of the Arbitrum Open House and BNB Chain Global Hackathon.
Core stack: TypeScript · Node.js · Python · Next.js · PostgreSQL · Docker · Azure/AWS · RAG · MCP · Agent Evaluation
I’m looking for contributors, early users, research collaborators, and design partners building enterprise agents, coding agents, context infrastructure, browser agents, secure tool execution, or governed automation.
Start in a project’s issues or discussions, or reach me at ajnas@cognifyr.co.