Integrated and agentic governance
Saidot is available in three modes: as a web application, as a REST API, and as a set of MCP (Model Context Protocol) servers. This section is for teams and practitioners who want to use Saidot beyond the UI — to embed governance into AI development workflows, automate routine actions, or connect AI assistants directly to governance data.
If you are new to Saidot and want to set up your organisation, register your first system, or manage risks through the interface, start with Getting started with Saidot instead.
Who this section is for
Governance leads who want to automate repetitive tasks or connect Saidot to other tools in the organisation
Developers and platform engineers building integrations — from CI/CD hooks and deployment pipelines to event-driven governance workflows
AI assistants and agents (such as Claude, GitHub Copilot, or custom-built agents) that read and write governance data programmatically on behalf of users
What's in this section
How Saidot governance works
The conceptual data model: what entities exist in Saidot (Systems, Risks, Controls, Policies, Models, Datasets, Agents), how they relate to each other, and the distinction between the curated Library and your organisation's own Governance data. Start here before working with any API or MCP tool.
Governance workflows by role
Role-by-role guide to what governance actions look like when triggered programmatically - for AI governance leads, risk and compliance teams, system owners, developers, and reviewers. Also documents the pre-built workflow Skills delivered via GitHub, so your team can adopt them directly or use them as starting points.
Event-driven governance patterns
Design patterns for integrating Saidot into automated pipelines: new model deployment → register in Saidot → risk inheritance fires → owner notified. Covers CI/CD hooks, deployment events, incident triggers, and scheduled audit runs. Read the Webhooks and Incidents articles first.
Connecting AI tools to Saidot
Step-by-step setup for Claude Desktop, VS Code, GitHub Copilot, and custom agents. Includes the canonical system prompt template and a worked conversation example showing how an agent navigates the governance data.
Governing agents you use for AI governance with Saidot
We recommend you to keep record and govern your AI governance agents in a similar way as other agents and AI systems. Using AI to automate and streamline AI governance processes can bring significant benefits, but also carry risks. Human oversight and responsibility is required to evaluate and approve the outcomes, mitigate risks and make required approvals and compliance assessments.
Before you start
You will need either:
For API access: A Saidot API token - generate one at Admin → API tokens.
For MCP access: The URLs of the relevant MCP servers - no API token required. Authentication is handled via OAuth when you connect. Note that MCP must be enabled for your organisation, and the first connection requires an administrator to approve it.
Library vs. Governance data
The Library and your Governance data are separate stores. The Library is Saidot's curated, read-only content. Your Governance space is where your organisation's systems, risks, and controls live and can be modified. Agents must understand which store to read from and which to write to. See How Saidot governance works for a full explanation.