Redberry Labs prices AI insurance at the agent level. Rather than issuing a single policy against your organisation as a whole, the underwriting engine evaluates every agent you deploy — examining its permissions, tools, data access, and guardrails — and assigns each one a risk score before calculating a premium. The result is coverage that reflects the actual risk each agent carries, not a blunt average across your entire estate.Documentation Index
Fetch the complete documentation index at: https://redberrylabs.com/docs/llms.txt
Use this file to discover all available pages before exploring further.
Why agent-level underwriting matters
Two agents can run on the same underlying model and still carry very different risk profiles. An agent that drafts internal summaries operates with read-only access to a narrow data set; its blast radius if something goes wrong is limited. An agent that approves payment transfers holds write access to financial systems and acts without a human in the loop for every transaction. Grouping them under a single policy would misprice both — overcharging the low-risk agent and undercharging the high-risk one. Redberry Labs underwrites each agent separately so that the premium matches the specific permissions, tools, users, and controls in place for that agent.How underwriting works
Submit your agents
In the Redberry Labs app, you provide details for each agent: its deployment context, the system prompt it runs on, the tools it can call, the users or systems it interacts with, and the volume of activity you expect.
Controls check
The underwriting engine inspects the controls you have in place. It evaluates permissioning (whether the agent follows least-privilege principles), tool safety, human-in-the-loop gates, data minimisation practices, auditability, and model and prompt governance.
Risk scoring
Each agent is scored across five risk dimensions: Misrepresentation, Operational Failure, Financial Error, Data Exposure, and Regulatory Breach. The score reflects the combination of the agent’s capabilities and the strength of the controls you have implemented.
Loss modelling
The algorithmic underwriting engine models potential loss scenarios for the agent based on its risk score, deployment scale, and the coverage types that apply. This produces an expected loss estimate that feeds directly into the premium calculation.
Policy structure
Your organisation holds a single base policy with Redberry Labs. Each AI agent you insure is scheduled onto that policy individually. The schedule records the agent’s name, deployment context, coverage types, limits, and the controls that were in place at the time of underwriting. This structure means you can add agents, remove agents, or adjust coverage for a single agent without rewriting your entire policy. It also provides a clear audit trail — you can see exactly what was covered, at what limit, and under what conditions, for every agent in your estate.The Redberry Labs SDK monitors your agents’ configuration continuously. If an agent’s permissions, tools, or system prompt change materially, the SDK flags the change so you can submit an updated evaluation. Coverage may be affected if changes are not disclosed.