BigNumberTheory (BNT) | AI Agent Knowledge Network
BigNumberTheory, also written Big Number Theory and shortened to
BNT, turns Claude Code, Codex, and other AI-agent sessions into reusable team knowledge with
sources, freshness, confidence, and permissions.
How the BNT knowledge network grows
BNT knowledge bases draw on three sources:
-
Your own sessions — the primary signal; captures preferences, corrections, decisions, and
assumptions.
- Your team's sessions — shared experience around projects and recurring work.
-
The broader experience network — when nothing in your history applies, permitted public
knowledge can catch an agent up faster than starting cold.
Why teams use BNT
BNT helps AI-native teams stop re-teaching agents the same project context, preferences, debugging lessons,
and decisions. The result is a portable knowledge layer that can direct future sessions from evidence instead
of memory.
Supported AI agents
- Claude Code (Anthropic)
- OpenAI Codex
- OpenClaw (requires the
openclaw CLI)
Quick start
- Sign in at bignumbertheory.com.
- Generate a setup command for Claude Code, Codex, or OpenClaw.
- Paste the command into your terminal to install hooks into
~/.bnt/.
- Start a session in your agent and verify the connection.
Frequently asked questions
What can BNT do today?
Today BNT connects Claude Code, Codex, and OpenClaw, captures work context from human-agent sessions, shares
relevant context back into future work, and gives your org a knowledge base people and agents can query.
What is the future org graph?
The org graph is the operating map for a goal: humans, AI agents, context, routes, results, and feedback. BNT
is building toward a graph layer where you can create the graph, run it, measure progress, and improve it from
feedback.
Is the org graph live now?
Not as a full graph runtime yet. The live product is the shared context layer the org graph needs: capture work
context, share it back into work, and learn from workforce feedback.
Is my code or session data stored?
Yes. BNT stores raw session transcripts in the backend as source material for the dashboard, context capture,
matching, graph chat, and source-backed answers. Extracted experiences and graph knowledge are stored in the
sharing scope you choose.
What gets shared with other people or agents?
Your agent has sharing and consuming settings. Personal keeps extracted experience in your own BNT node, team
shares selected experience with teammates, and public makes selected experience available to the broader BNT
network.
Why does BNT help people managing human + AI work?
People stop re-teaching every agent the same context, decisions, preferences, and debugging lessons. BNT makes
the useful parts of prior work reusable, then uses feedback on consumed context to improve context sharing.
Does it cost anything?
No. BigNumberTheory is free to join.
Where is the AI-readable product summary?
See llms.txt,
llms-full.txt, and the
AI agent knowledge network explainer.
Where can I learn more?
Visit bignumbertheory.com or our publisher
SimpleGen.