NullRun
NullRun is a runtime authorization layer for AI agents. It sits between AI agents and external tools, APIs, or services to enforce governance policies before actions are executed. The platform is designed to control agent spend, block unsafe tool calls, require human approval for sensitive actions, and maintain an audit trail of every decision. It is delivered as a paid API service and can be placed in front of agent frameworks. The product targets AI agent governance, cost control, safety, and compliance. According to its homepage, it supports integration with common agent frameworks and model providers, including LangChain, LangGraph, CrewAI, OpenAI, Anthropic Claude, Google Gemini, Cohere, and Mistral. It also references Model Context Protocol (MCP) authorization, fail-closed enforcement, API key rotation, and SOC 2-related compliance use cases. NullRun's core value is enforcing runtime policies that traditional observability or post-hoc logging tools do not provide. It can cap spend, filter tool calls, route approval requests to humans, and record an auditable history for security and compliance teams. The service is aimed at teams deploying autonomous agents in production environments where uncontrolled actions or costs present operational risk.
- Category
- Apis
- Pricing
- Paid
Problem solved
AI agents can autonomously call tools and spend money, but without runtime controls they may execute unsafe actions, exceed budgets, or leave no auditable record. NullRun solves this by acting as an authorization checkpoint between agents and external systems, enforcing spend limits, blocking or gating risky tool calls, and logging every decision for governance and compliance.
Use cases
- Enforce real-time spend limits and budget caps for LLM API calls made by AI agents.
- Block or require human approval for unsafe or high-risk tool calls before execution.
- Add human-in-the-loop approval workflows for sensitive agent actions.
- Maintain an audit log of AI agent decisions and tool calls for compliance (e.g., SOC 2).
- Integrate runtime authorization with LangChain, LangGraph, CrewAI, or MCP-based agents.
- Govern API keys and rotate credentials for model provider access.
- Apply fail-closed enforcement and rate limiting to agent API usage.
Audience
- Developers
- ML Engineers
- Enterprise
- Startups
Pros
- Runtime enforcement for AI agent tool calls and spend, not just post-hoc observability.
- Supports human-in-the-loop approvals for high-risk agent actions.
- Provides audit trails for governance and compliance use cases.
- Designed to integrate with multiple agent frameworks and model providers.
- Includes fail-closed enforcement and API key rotation for security.
Cons
- Paid product; no free tier or open-source edition is indicated on the homepage.
- Requires integration into existing agent frameworks and may need code changes.
- Public technical documentation is limited in the available source material.
- Effectiveness depends on correctly configuring policies and approval rules.
- No public GitHub repository or community edition was available for review.
Official website: https://nullrun.io/
Canonical LambdaBase page: https://www.lambdabase.com/tools/nullrun