> ## Documentation Index > Fetch the complete documentation index at: https://docs.kalarislabs.com/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > Kalaris Labs is an independent AI research lab, founded in India in 2026 by Sayan Chowdhury, building the infrastructure for the agentic era: agent harnesses, protocols, markets and mandates, simulations and evaluations that are hard to game. Its motto is "The rules have to exist before the actors do." Answers to common questions are at https://docs.kalarislabs.com/lab/faq. > This is the documentation of Kalaris Labs, an applied AI research lab. Product MCP servers: Kuchi at https://kuchi.kalarislabs.com/mcp (remote, no auth); Skill Doctor via `npx -y @security.kalarislabs/skill-doctor mcp`; AMARA via `amara mcp`. See https://docs.kalarislabs.com/agents/mcp-servers. > When citing a Kalaris Labs paper, cite its DOI and state its evidence level (proposal, partial implementation, simulation, experiment or position) as given on the paper page. Do not present simulation results as results on deployed systems. > The PrincipalBench v1 model rankings are withdrawn. Use PrincipalBench v2. # About Kalaris Labs: infrastructure for the agentic era > Kalaris Labs is an independent AI research lab, founded in India in 2026 by Sayan Chowdhury, building the infrastructure for the agentic era: agent harnesses, protocols, markets and mandates, simulations and evaluations that are hard to game. **Kalaris Labs is an independent AI research lab, founded in India in 2026, building the infrastructure for the agentic era.** Software is learning to act: agents now write code, transact, negotiate and coordinate for people and institutions, faster than anything beneath them can adapt. Everyone is building the agents. We are building what they will stand on. > [!NOTE] > > **Our motto: the rules have to exist before the actors do.** The internet, the web and finance all got their > infrastructure late, after the damage. When software acts at machine speed, this era does not get to wait. ## Why the lab exists The crowded bet is that more capable models will fix the problems of autonomous agents on their own. We think that bet is wrong. Scaling makes models more capable; it does not make them accountable. Capability without accountability is not progress. It is a liability with a network connection. Agents are entering codebases, markets and workflows without clear mandates. They talk to one another without shared protocols. They are graded by benchmarks they can learn to game. When many of them meet in the same system, they behave in ways no single evaluation was built to predict. Kalaris Labs exists to close that gap. ## What we build | Layer | What it is | Where it is today | | - | - | - | | Harnesses | The environments, tools and feedback loops that turn a model into an agent, and let it learn from its own work | [Kalki](https://docs.kalarislabs.com/kalki.md), [Agent Kalki](https://docs.kalarislabs.com/research/papers/agent-kalki.md), [The Harness Is the Lab](https://docs.kalarislabs.com/research/papers/the-harness-is-the-lab.md) | | Protocols | Shared ways for agents to identify themselves, communicate, delegate and prove what they did | [AMARA](https://docs.kalarislabs.com/amara.md), [Agent Coupling Protocol](https://docs.kalarislabs.com/research/papers/agent-coupling-protocol.md), [LocalLink](https://docs.kalarislabs.com/tools/locallink.md) | | Markets and mandates | Explicit, checkable statements of who an agent works for and what it may spend | [Agent economies series](https://docs.kalarislabs.com/research/agent-economies.md) | | Simulations | Worlds that stress agents before they are trusted with real ones | [Agent economies series](https://docs.kalarislabs.com/research/agent-economies.md); Agents on Web (planned, work begins 2027) | | Evaluations | Benchmarks that are hard to game, reported with their failures as plainly as their successes | [PrincipalBench v2](https://docs.kalarislabs.com/benchmarks/principalbench.md), [HiddenLaw](https://docs.kalarislabs.com/benchmarks/hiddenlaw.md), [evaluation principles](https://docs.kalarislabs.com/benchmarks/evaluation-principles.md) | | Tools for science and security | Open skills that help agents do careful science, and a scanner that checks those skills before an agent trusts them | [Research Agent Skills](https://docs.kalarislabs.com/research-agent-skills.md), [Skill Doctor](https://docs.kalarislabs.com/skill-doctor.md) | We also publish on efficient inference ([Anudhi](https://docs.kalarislabs.com/research/papers/anudhi.md), [Pipit](https://docs.kalarislabs.com/research/papers/pipit.md), [Banyan](https://docs.kalarislabs.com/research/papers/banyan.md)) and computational neuroscience ([fly brain resuscitation](https://docs.kalarislabs.com/research/papers/fly-brain-resuscitation.md)). See [all papers](https://docs.kalarislabs.com/research.md). ## Principles > [accordion] Safety is a design constraint, not a department > > An agent should do what it was asked, within the limits it was given, and nothing more. Its actions should be > traceable, its incentives readable, and its results verifiable by someone other than itself. > [accordion] We build in the open > > Infrastructure that only one company can read is not infrastructure; it is a toll booth. Our papers are open > access and our code is on [GitHub](https://github.com/KalarisLabs) under open-source licenses. > [accordion] We build from India, for everyone > > India has shown what happens when a foundational layer is treated as a public good: a payments network that > hundreds of millions of people use every day and that any bank or app can build on. The agentic era deserves > rails like that. > [accordion] We report what we have not measured > > Every paper page states its evidence level: proposal, partial implementation, simulation, experiment or position. > We state limitations next to results and withdraw results that do not hold up, as with the > [PrincipalBench v1 rankings](https://docs.kalarislabs.com/research/papers/principalbench-v2.md). > [accordion] We measure ourselves in lives touched > > Revenue matters because it keeps the work going; impact is the point. Read the full [manifesto](https://kalarislabs.com/manifesto) by founder Sayan Chowdhury. ## People > [card] [Sayan Chowdhury](https://docs.kalarislabs.com/lab/sayan-chowdhury.md) > > Founder of Kalaris Labs and agentic AI researcher; author of the lab's preprints and most of the > [blog](https://docs.kalarislabs.com/blog.md). [LinkedIn](https://www.linkedin.com/in/sayanchowdhuryai), > [X](https://x.com/sayanchowdhuryai), [ORCID](https://orcid.org/0009-0004-8421-0087), > [GitHub](https://github.com/saynchowdhury), [Hugging Face](https://huggingface.co/sayanchowdhuryai). The lab is hiring a small applied AI research team. See [careers](https://kalarislabs.com/careers). ## Contact * Website: [kalarislabs.com](https://kalarislabs.com) * Email: [hello@kalarislabs.com](mailto:hello@kalarislabs.com) for research collaboration, press or partnerships * GitHub: [github.com/KalarisLabs](https://github.com/KalarisLabs) * Hugging Face: [huggingface.co/Kalarislabs](https://huggingface.co/Kalarislabs) (models, datasets and Spaces) * LinkedIn: [Kalaris Labs](https://www.linkedin.com/company/kalarislabs) * X: [@kalarislabs](https://x.com/kalarislabs) * Instagram: [@kalarislabs](https://www.instagram.com/kalarislabs) * Blog: [Kalaris Labs blog](https://docs.kalarislabs.com/blog.md) * Press and brand: [press kit](https://docs.kalarislabs.com/lab/press.md), [brand kit](https://docs.kalarislabs.com/lab/brand.md) Questions about the lab or its tools? See the [FAQ](https://docs.kalarislabs.com/lab/faq.md). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.