Sponsio
View on GitHubDeterministic safety solutions for probabilistic AI agents
A Python library and CLI that checks AI agent tool calls against runtime contracts before execution. Its deterministic, low-latency rules support common agent frameworks and MCP, with actions to pass, block, warn, escalate, or redirect.
Use Cases
Block destructive tool calls before executionEnforce approval requirements for sensitive actionsPrevent unauthorized refunds or financial transactionsApply runtime policies to shell and filesystem toolsAudit agent actions and escalation decisionsGuard multi-tool workflows against unsafe action sequences
Built With
- Language
- Python
- Frameworks
- LangGraph · CrewAI · OpenAI Agents SDK · Claude Agent SDK · Google ADK · Vercel AI SDK · MCP
Tags
agent-safety · runtime-guardrails · agent-security · contract-enforcement · tool-call-verification · policy-engine · prompt-injection · deterministic · audit-trails · self-hosted