Contracts
AgentAssert
Design-by-contract for agents: preconditions, invariants, governance and recovery checked at runtime, so a violation is caught at the step that caused it.
CodeAGPL-3.0
Topic hub · Contracts
Prompts state intent. Contracts enforce it.
A prompt asks an agent to behave; nothing checks that it did. Contracts state preconditions, invariants, governance policies and recovery, check them at runtime, and give each decision a typed answer and a receipt.
Start here
Read it in this order.
Your progress is saved in this browser.
Reading path · 0 of 6 read
- PaperAgent Behavioral Contracts: Formal Specification and Runtime Enforcement for Reliable Autonomous AI Agents
- ToolAgentAssert: Behavioral Contracts for AI Agents
- PostJev Decision Layer: Save Frontier Tokens on Closed Decisions
- PostJev for Codex: Introducing Qualixar Jev Control v1.1.1 and Policy Mode
- PostAI Agent Observability Is Not Enough. You Need an Evidence Plane.
- PaperAgent Behavioral Contracts II: Certifying Compositional Reliability Without Assuming Independence
Tools
What to run.
Contracts
AgentAssert Type-C
Behavioural contracts with no agent code changes: an HTTP proxy that enforces budgets, latency, tool allowlists, topic bounds and PII filters.
CodeMIT
Decisions
Jev Decision Layer
Typed task, tool, skill and review routing for agents, with local policy gates and an auditable receipt for every decision.
CodeMIT
Research
The papers.
Agent Behavioral Contracts: Formal Specification and Runtime Enforcement for Reliable Autonomous AI Agents
Design-by-contract for agents — invariants that hold on every execution path. 1,980 sessions across 7 models
Agent Behavioral Contracts II: Certifying Compositional Reliability Without Assuming Independence
Two instances of the same model failed together on 90% of missions where either failed — independence is not a safe assumption
Glossary
Terms you'll meet.
- Behavioural contract
Design-by-contract applied to an autonomous agent: formal, runtime-enforceable rules about what it must and must not do, with recovery when it drifts.
- Contract components (P, I, G, R)
The four parts of an agent behavioural contract: preconditions that must hold before the agent acts, invariants that must hold throughout, governance policies, and recovery mechanisms for when a constraint is broken.
- Failure independence assumption
The common but rarely tested assumption that agents in a system fail independently. Measured dependence is strongly positive, so bounds built on it understate joint failure.
- Contract proxy
An HTTP proxy between an agent and its model provider that enforces contract rules, such as budgets and tool allowlists, with no change to the agent's code.
- Typed decision
A bounded choice, such as which tool or skill to use, answered from a fixed set of options with a confidence and a local receipt. The host keeps execution authority.
Writing
Posts on contracts.
What's changing
Jev Decision Layer: Save Frontier Tokens on Closed Decisions
Send closed agent choices to TypeSafe Jev at $0.042 per million input tokens, with output free. Qualixar adds the local gate and receipt. The host still acts.
25 min read
What's changing
Jev for Codex: Introducing Qualixar Jev Control v1.1.1 and Policy Mode
Qualixar Jev Control v1.1.1 adds TypeSafe Jev System One decisions and selective Policy Mode to Codex for routing, ranking, context selection, evidence checks and more.
18 min read
How we built it
AI Agent Observability Is Not Enough. You Need an Evidence Plane.
A reference architecture for proving what an AI agent ran, what it read, what it was allowed to do, and why its output should be trusted.
14 min read
Research explained
The First Token Knows — and Where That's Not Enough
Why first-token entropy is the cheapest hallucination signal in production — and the layer of runtime contracts and statistical assays that needs to sit on top.
15 min read
Research explained
Stop Prompting. Start Contracting. Why 15% of 'Never Delete User Data' Prompts Fail — and What Replaces Them.
A viral Reddit thread proved agents ignore safety prompts in 15% of edge cases. Gartner says 40% of agent projects die by 2027. The fix isn't a better prompt — it's a runtime contract. Here's the AgentAssert + AgentAssay playbook.
8 min read
How we built it
Your AI Agent Passed Staging. Then It Hallucinated a Migration in Production.
Your test suite is green. Every unit test passes. Integration tests pass. The agent generates correct...
3 min read