# CIFR — Council for Independent Frontier Research

> Research should run, not just read.

CIFR is a registered nonprofit (501(c)(3) in the USA; Section 8, Section 12, 80G tax-deductible, CSR-eligible in India) pre-publication research validation platform that makes academic research reproducible, executable, and composable. Researchers submit their code, data, and methods — CIFR runs them in isolated containers and produces citable, reusable research agents.

## What CIFR Does

- Turns published research papers into callable, executable **Research Agents**
- Each agent runs the paper's original code in an isolated Docker container with pinned dependencies
- Outputs are captured as verifiable artifacts with full provenance tracking
- Agents can be composed together — one researcher's method becomes another's building block
- Every invocation is logged, giving authors real-time impact metrics beyond citation counts

## Research Agent Identifier (RAI)

CIFR issues the world's first **Research Agent Identifier (RAI)** — a permanent, unique identifier for executable research methods. If a DOI tells the world *what* you published, an RAI tells the world *what your research can do*.

Format: `RAI:cifr/<author>-<method>-<year>/v<version>`

Example: `RAI:cifr/chanda-resiliency-2016/v1`

## How It Works

1. **Submit**: Paste a GitHub URL, upload a ZIP, or drag in a Jupyter notebook
2. **Execute**: Code runs in a clean, isolated container with auto-detected dependencies and no network access (for reproducibility)
3. **Publish**: Your paper becomes a living, callable agent that anyone can invoke, build on, and cite

## Key Features

- **Free for academics**: No paywalls, no data harvesting, no vendor lock-in. Registered nonprofit in both USA (501(c)(3)) and India (Section 8, 80G, CSR-eligible)
- **Reproducible execution**: Isolated Docker containers with pinned environments
- **Real-time impact tracking**: Know when someone builds on your work, not just cites it
- **Composable research**: Chain agents together into multi-step research pipelines via a visual DAG editor
- **Provenance hashing**: Every run produces a verifiable provenance hash
- **Multiple provenance types**: Author-original, reference implementations, independent reproductions
- **Trust system**: Agents earn trust tiers based on verification status and community validation
- **Python SDK**: Programmatically invoke agents from your own code
- **MCP integration**: Call research agents from AI assistants via Model Context Protocol

## For Institutions

CIFR offers institutional memberships for universities, research labs, and companies:

- **SOC 2 compliant platform** for secure research execution
- **GPU compute hours** for accelerated model training and inference
- **Cloud Labs**: Department-level or organization-level isolated research environments
- **Auto-Research pipelines**: Automated experiment execution as new data arrives
- **Agent marketplace access**: Invoke any published agent within institutional workflows
- **Bulk researcher onboarding**: Streamlined access for all team members
- **Priority support and SLAs**

## Supported Languages and Environments

- Python (requirements.txt, pyproject.toml, Pipfile, environment.yml)
- Node.js (package.json)
- Any language with a Dockerfile
- Jupyter notebooks (.ipynb)
- R and Julia (coming soon)

## Links

- Website: https://cifr.org.in
- Living Archive (browse agents): https://cifr.org.in/agents
- Submit an experiment: https://cifr.org.in/agents/new
- Documentation: https://cifr.org.in/docs
- Compose agents: https://cifr.org.in/compose
- Institutional memberships: https://cifr.org.in/institutions
- API reference: https://cifr.org.in/docs/api
- Python SDK guide: https://cifr.org.in/docs/python-sdk
- Contact: contact@cifr.org.in

## Docs

- [Quickstart](https://cifr.org.in/docs/01-quickstart)
- [cifr.yml Reference](https://cifr.org.in/docs/02-cifr-yml-reference)
- [Research Agent Identifiers](https://cifr.org.in/docs/03-research-agent-identifiers)
- [Python SDK](https://cifr.org.in/docs/04-python-sdk)
- [Composition Runtime](https://cifr.org.in/docs/05-composition-runtime)
- [Trust System](https://cifr.org.in/docs/06-trust-system)
- [Provenance Types](https://cifr.org.in/docs/07-provenance-types)

## Optional

- [Full LLMs.txt with extended detail](https://cifr.org.in/llms-full.txt)