Sign in Get started
Now in early access

The trust layer for AI agents

Governance, observability, and policy enforcement for every AI agent in your organisation. Deploy with confidence.

Request early access Open dashboard →
Compatible with leading agent frameworks
LangChain OpenAI Anthropic CrewAI AutoGen Cohere Google Cloud AWS Bedrock Mistral Hugging Face Salesforce Microsoft ElevenLabs LangChain OpenAI Anthropic CrewAI AutoGen Cohere Google Cloud AWS Bedrock Mistral Hugging Face Salesforce Microsoft ElevenLabs

Belt and braces governance for every agent

Choose how deeply Spud integrates. From lightweight observability to full dual-enforcement — one SDK, three modes.

Mode 01
👁️

Observe

Monitor agent activity in real-time with zero performance overhead. Every decision, tool call, and data flow — fully visible.

Zero latency impact
Mode 02
🛡️

Govern

Enforce policies inline. Evaluate every action against your rules and approve, modify, or block before it reaches the real world.

Inline enforcement
Mode 03
🔐

Enforce

Full dual-enforcement architecture. Policies run in the SDK and at the gateway — compliance even if one layer is bypassed.

Dual enforcement

Everything you need to trust your agents

Privacy first

Privacy-preserving evaluation

Evaluation runs locally within your trust boundary. Customer data never leaves your infrastructure — models trained on synthetic data, deployed alongside your agents.

policy: financial-data-handling
evaluation: local
data_egress: false
model: spud-eval-v2-local
rules:
  - block_pii_in_tool_calls: true
  - require_approval_over: $10,000
  - audit_log: immutable
Performance

Built for speed

Governance that doesn't slow your agents down.

<2ms
Local evaluation latency
100%
Audit trail coverage
Developer experience

One SDK, any framework

Wrap any LLM call or agent framework with a single import. LangChain, CrewAI, AutoGen, custom builds.

import { SpudGovernor } from '@spud/sdk'

const gov = new SpudGovernor({
  mode: 'govern',
  policies: ['financial', 'pii'],
  localEval: true
})

// That's it. Wrap and go.
const agent = gov.wrap(myAgent)
Control plane

Governance dashboard

A unified control plane for every agent in your org. Policies, incidents, approvals, and compliance — all in one place.

Agents
47
Policies
12
Blocked
3
Compliance
99.7%

Governance that speaks agent

Native support for the protocols powering the next generation of AI.

MCP

Model Context Protocol

Govern tool calls, context injection, and resource access for agents using MCP servers — full visibility into data flows.

Tool-level policiesContext auditingResource gates
A2A

Agent-to-Agent Protocol

Enforce governance at agent boundaries. Trust relationships, capability delegation, and cross-agent audit trails.

Trust boundariesDelegation rulesCross-agent audit
🔒

SOC 2 Ready

Built with enterprise compliance from day one.

🏠

Data Stays Home

Privacy-preserving local evaluation. Your data, your infrastructure.

📋

Immutable Audit

Every agent action logged with tamper-proof audit trails.

Zero Overhead

Sub-2ms evaluation. Your agents stay fast.

Deploy agents
you can trust

Join the teams building the future of AI agent governance. Early access is open now.