Persistent threads
Come back to the same thread with its context, sources, and findings still in place, so your research can keep building instead of restarting from scratch.
Harnesses
Pre-built AI agents that deliver specific outcomes — research portfolios, production monitors, engineering pipelines. One prompt, one deliverable, no plumbing.
Research harness
The one harness you can buy right now. It turns a question into a collection that keeps growing instead of a chat you have to re-run every time you come back.
Come back to the same thread with its context, sources, and findings still in place, so your research can keep building instead of restarting from scratch.
Find and filter the sources behind your question across thousands of sites, then keep the useful evidence in the same research thread.
Use your own API key to run research and keep the setup connected to the tools you already use.
Open source
Build your own agents and harnesses through our harness primitive SDK. Every primitive beneath a Vidbyte harness is public.
One install. The full Vidbyte stack: agents, tools, middleware, context, evals, pipelines, and tracing, packaged as an MIT-licensed Python SDK with no Vidbyte account required.
One command. Control Vidbyte from your terminal: create jobs, manage API keys, and interact with harnesses through one focused CLI entry point with no separate dashboard required.
One fetch. Give your agent the SDK and CLI context it needs: fetch the public llms.txt guides once, then work with every Vidbyte surface without guessing at commands or boundaries.
Coming soon Set up API access from an agent: create a key, verify the connection, and fund usage on Vidbyte through a guided skill that keeps credentials out of the conversation.
All packages are MIT licensed and require Python 3.11+.
Vidbyte
Harnesses that do a specific job, keep the evidence, and stop at the budget you approved.
Every harness draws on one Vidbyte account and one credit balance.