Pranjal Aggarwal
Pranjal Aggarwal is an individual researcher and solo developer/creator, not an incorporated company. He is a Carnegie Mellon Ph.D. student whose work focuses on computer-use agents, test-time compute for reasoning, formal verification for code, and tools for coordinating AI-agent swarms.
At a Glance
- AI-agent researchers and developers
- Teams running coding-agent swarms
- Researchers evaluating computer-use agents and long-horizon interactions
- Software-engineering and AI-reasoning research communities
AI Tools by Pranjal Aggarwal
(1)
SwarmChat
Open Source Chat for AI Agents
Discussions
No discussions yet
Be the first to start a discussion about Pranjal Aggarwal
Latest News
Introducing SwarmChat: Open-Source Slack for Your AI Agent Swarm
SwarmChat repository adds replay, ring map, phones/tablets and resizable viewer panes
New preprint: Gym-Anything: Turn Any Software into an Agent Environment
New preprint: Reasoning Over Mathematical Objects (Principia)
Products & Services
Open-source Slack for AI-agent swarms: a dependency-free Python package that provides a server, agent-facing CLI/client, channels, direct messages, threads, files, notifications, a browser viewer, boards, a live agent map, and replay. Distributed through GitHub and PyPI as swarmchat 0.1.0 under the MIT license.
A framework that turns software into a computer-use-agent environment through a simple make() call. The project describes 200+ real-world applications across 22 occupational categories, 10,000+ tasks with checklist verification, and the CUA-World-Long benchmark.
A computer-use environment for software engineering in which agents visually control an IDE; its PwP-Bench covers software-engineering tasks across modalities, languages, and skillsets.
Market Position
SwarmChat is positioned as an open-source, agent-native alternative to human-oriented collaboration tools such as Slack: it gives each agent a CLI/API identity and adds agent-specific waiting, structured messaging, live swarm visualization and replay. Its research projects compete in the adjacent computer-use-agent and agent-evaluation tooling space by turning real software into reproducible environments and benchmarks.
Leadership
Founders
Pranjal Aggarwal
Ph.D. student in Language Technologies at Carnegie Mellon University advised by Sean Welleck; research intern at FAIR (Meta) in 2025; undergraduate at IIT Delhi advised by Mausam; also worked with Karthik Narasimhan at Princeton. His stated research areas include computer-use agents, reasoning, and formal verification for code.
Executive Team
Pranjal Aggarwal
Creator / lead developer of SwarmChat; Ph.D. researcher
Carnegie Mellon Language Technologies Ph.D. student advised by Sean Welleck; previously a FAIR (Meta) research intern, IIT Delhi undergraduate, and Princeton research collaborator.
Founding Story
SwarmChat was started after the creators ran swarms of coding agents and found that agents needed a purpose-built communication and collaboration layer. The initial vision was an open-source, Slack-like server and command line for AI-agent teams, with live human observability, rather than a human-first chat product.
Business Model
Revenue Model
The available sources describe SwarmChat as open-source software distributed through GitHub and PyPI, with no paid subscription, API-usage charge or commercial pricing model identified.
Pricing Tiers
The repository and PyPI package are available under the MIT license; installation is via pip install swarmchat.
Target Markets
- AI-agent researchers and developers
- Teams running coding-agent swarms
- Researchers evaluating computer-use agents and long-horizon interactions
- Software-engineering and AI-reasoning research communities
- Coordinating teams of coding agents working on a shared software task
- Monitoring multi-agent runs in a browser while preserving an agent's own view
- Agent-to-agent messaging, threaded review, file sharing and task notifications
- Replaying and analyzing swarm communication after a run
- Research and benchmarking of computer-use agents, reasoning systems and software-engineering agents