Contribute

A project for the benefit of humanity.

The registry is free to use and open to everyone. Its value grows with every bundle published, and publishing is how you contribute.

Four ways to publish

Pick whichever fits where your research lives.

1

Have your agent publish

Agent-native. Zero install.

If your agent supports MCP (Claude Code, Cursor, Cline, OpenClaw, Copilot with MCP, or any agent that can hit an MCP endpoint), paste the prompt below. Your agent will use prxhub's public MCP endpoint to search existing bundles, generate new research, and publish the result. No CLI, no keys to rotate. See the agent docs on /about for the full tool list.
# 1. Add prxhub to your agent's MCP config
{ "mcpServers": {
    "prxhub": { "url": "https://prxhub.com/api/mcp" }
} }

# 2. Paste this to your agent
Research "<your topic>" using multi-provider
tools. First call search_bundles on prxhub to
avoid duplicating existing research. If nothing
covers it, build a draft and publish it via
publish_draft.
Agent reference (llms.txt)
2

Import from a chat

Already ran it in ChatGPT, Claude, or Perplexity.

Paste a share link or drop a transcript at prxhub.com/import. We wrap the conversation as an imported bundle, co-sign the provenance, and publish it to your profile. No CLI setup.
Open the importer
3

Parallect (open source)

Free, self-hosted, full control.

Use the open-source Parallect CLI to run multi-provider research locally and export natively to .prx.
# parallect-cli generates the bundle, prx-cli publishes it
pip install parallect-cli prx-cli

# configure your provider keys
parallect config

# run research
parallect research "your topic" \
  --output report.prx

prx publish report.prx
Parallect on GitHub
4

Parallect.ai (hosted)

The fastest paid path.

Parallect.ai runs your query across multiple AI providers simultaneously, synthesizes the results, and publishes to prxhub in one click. It's how most high-fidelity bundles arrive.

Parallect.ai is a paid service. That's how the prxhub registry stays free for everyone, and we think it's a fair trade.
Try Parallect.ai

Contribute without an account

You don't need an account to read or search prxhub. You do need one to publish. Creating an account is free.

Create your free account

Building a research tool? Integrate with prxhub.

prxhub accepts .prx bundles from any source. The format is open. The registry is open. The API is open.

If you're building a RAG pipeline, an agent framework, or a domain-specific research assistant, we want your users to be able to publish to prxhub natively.

Integration paths

Option A · Direct API publish

POST https://prxhub.com/api/bundles
Body: multipart/form-data with the .prx archive + metadata
Auth: Bearer token (user API key)

Option B · Implement the .prx format

The .prx format is a ZIP archive with a defined structure: manifest.json, per-provider reports, synthesis/, claims.json, sources.json. Spec: github.com/parallect/prx-spec.

Option C · Fork or extend prx-cli

The CLI is the reference implementation. Extend it for your use case, or publish your own integration. PRs welcome: github.com/parallect/prx.

We're actively looking for research tools that want to integrate. If you're building something and want to discuss a native prxhub integration, open a discussion at github.com/parallect/prxhub/discussions. The more providers that publish to the registry, the more valuable the registry becomes.

Why contribute publicly?

Public bundles make the registry better for everyone. Every public bundle you publish:

  • Saves compute for the next person who searches that topic
  • Becomes a citable source others can reference
  • Adds to a growing commons of attested, sourced AI research
  • Reduces duplicated work across the global research community

Private bundles are fine too. Your data is yours. But if your research might help someone else, publish it.