# Agentic Thinking - Full Site Content for LLMs

## Company Overview
Agentic Thinking Limited is a UK independent research lab. We test AI agent runtimes and the LLMs behind them on real tasks, reconstruct agent incidents from evidence, and publish what we find: we test what AI agents really do, and investigate when it goes wrong. The questions we study are what an agent did, why, and who approved it, and whether a record of that can be trusted by someone other than its author.

## Flagship: AgenticBench
AgenticBench is our flagship work. Independent, repeatable tests of what AI coding agents actually do on your machine. The scorecard is updated as new agents, versions and tests are added: every result links to its evidence, vendors hear about findings before publication, and corrections are dated. The test rig is public under Apache 2.0: https://github.com/agentic-thinking/agenticbench. Scorecard and reproduction instructions: https://agenticbench.org

## Research
We publish independent reconstructions and analysis of AI agent incidents, built from public primary sources with stated confidence. Our 2026 notes cover the OpenAI evaluation-agent breakout and the Hugging Face intrusion, an Australian Medicare portal breach, agents using a public wiki as a message board, and the reporting gaps between when incidents happened and when those affected were told.
In September 2026 we replayed synthetic incident scenarios through two HookBus builds. On the earlier release, governance fields needed for reconstruction were lost in transit (0 of 75 delivered); with the published fix, 75 of 75 arrived. This is a small synthetic test, not a finished result.
We disclose that we build open evidence tooling in this area.

## Standard: AgentHook
AgentHook is an open runtime evidence standard (Apache 2.0) that Agentic Thinking stewards. It defines a vendor-neutral format for agent lifecycle events, actions, decisions and approvals, so a record means the same thing across agent runtimes. Specification, conformance fixtures and papers: https://agenthook.org

## Open tool: HookBus
HookBus is an open-source (Apache 2.0), self-hosted event bus that carries AgentHook-style lifecycle events from agent runtimes to subscribers. Publishers exist for Claude Code, OpenAI Codex CLI, Sourcegraph Amp, OpenCode, Hermes and OpenClaw. https://github.com/agentic-thinking/hookbus

## Evaluations
If your product runs an AI agent that takes actions, through tools, APIs, shells or other systems, you can commission an independent evaluation: what the agent can really do, whether your guardrails stop it at runtime, and whether your logs prove what happened. Each package has a fixed scope and fee agreed up front; the fee does not depend on the result. The report is delivered to you; whether and how it is published is agreed before work starts. Not a certification or a penetration test, and never an endorsement. Offered through Agentic Evals: https://agenticevals.dev AgenticBench is separate: it never accepts funding from the agents it tests, and coding agents on the benchmark are not eligible for paid evaluations. Enquiries: partnerships@agenticthinking.uk. https://agenticthinking.uk/evaluations.html

## AgentAuditor
AgentAuditor is our hash-chained, tamper-evident evidence recorder. It is used in our research and is not yet published as open source.

## Software and patents
Agentic Thinking does not sell software or any agent-governance product. Its software, AgentProtect (a set of HookBus policy and approval subscribers), AgenticStudio and HookBus Agent (an agent runtime), is not currently offered. AgentHook and HookBus are open source; Agentic Thinking has committed to transferring the AgentHook standard to an independent foundation. Patent applications by the founder, Leo Ruocco: HookBus GB2608069.7 and CRE GB2604445.3 (pending, to be assigned to the company). Patent enquiries: partnerships@agenticthinking.uk. The company holds no third-party compliance certifications.

## Company Details
- Agentic Thinking Limited, registered in England & Wales, Company No. 17152930
- Office 9, Sycamore Trading Estate, Squires Gate, Blackpool FY4 3RL
- Founded by Leo Ruocco, with a background in UK financial services, insurance and defence
- Patent applications by the founder, Leo Ruocco: HookBus GB2608069.7 and CRE GB2604445.3 (pending, to be assigned to the company)
- Trade marks: HookBus®, AgentHook®

## Key URLs
- Homepage: https://agenticthinking.uk/
- Research: https://agenticthinking.uk/research/
- Blog: https://agenticthinking.uk/blog/
- AgentHook: https://agenthook.org
- HookBus: https://hookbus.com
- Evaluations (Agentic Evals): https://agenticevals.dev
- Security: https://agenticthinking.uk/security.html