---
title: Which integrations should an AI agent governance platform have on day one?
url: https://opencontroller.lyzr.com/blog/day-one-integrations-for-ai-agent-governance
published: 2026-11-14
updated: 2026-11-14
author: Lyzr
---

# Which integrations should an AI agent governance platform have on day one?

**On day one you need identity, your cloud accounts, your model providers and your main SaaS agent platforms. Those four tiers decide whether the platform can see your agents at all. Code hosts, ITSM, SIEM and GRC can follow in the first quarter. For each integration, ask what data comes back, because a logo on a slide is not coverage.**

Every vendor in this category shows a wall of logos. The useful question is narrower: which integrations must work in the first week for the platform to be worth deploying, and what must each one return? This checklist is part of our [AI agent governance platform evaluation checklist](/blog/ai-agent-governance-platform-evaluation-checklist).

## Which integrations are must-haves for day one?

These four tiers are the minimum. Without them the platform has a partial inventory and cannot attribute anything to a person.

| Tier | Integration | Must return to count as covered |
|---|---|---|
| 1. Identity | Your identity provider over OIDC or SAML, plus SCIM | SSO for admins; users and groups for owner assignment; deprovisioning events |
| 2. Clouds | AWS, Azure and Google Cloud accounts | Agents, model deployments and API keys per account, with resource IDs and creation dates |
| 3. Model providers | Direct provider accounts and enterprise workspaces (for example Claude Enterprise, OpenAI Enterprise) | Workspaces, projects, keys, users and usage by key or project |
| 4. SaaS agent platforms | The agent builders you already pay for (for example Salesforce, SAP, Databricks, HubSpot, ServiceNow) | Agents, their tools or actions, their run-as identity, and status |

If a vendor covers tier 2 but not tier 4, it will miss the agents business teams build inside SaaS. If it covers tier 4 but not tier 1, you will have agents with no owner.

## Which integrations can wait for the first quarter?

These make the platform useful to more teams. Plan them, but don't block the rollout on them.

1. **Code hosts (GitHub, GitLab, Bitbucket).** Must return repositories that import agent SDKs, MCP configuration files and AI actions in CI workflows. See [how to find AI agents in your code repositories](/blog/find-ai-agents-in-source-code).
2. **ITSM and CMDB.** Must create or update a record per production agent and route incidents to its support group.
3. **SIEM and observability.** Must export events and traces in a standard format. Ask for OpenTelemetry, ideally following the [GenAI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/), plus webhooks for alerts.
4. **GRC tool.** Must push control evidence (inventory counts, posture findings, eval results) into the controls you already test, not just a PDF export.
5. **Developer devices.** Must report coding agents, local MCP servers and skills on laptops, installed through MDM.
6. **MCP servers and tool registries.** Must list servers, tools and which agents call them, using the [Model Context Protocol](https://modelcontextprotocol.io/specification) rather than a proprietary format.

## How do you test that an integration is real?

Ask for each integration to be shown working against a test tenant during the pilot. Then check:

- **Read depth.** Does it list only the agent, or also its tools, model, identity and versions?
- **Freshness.** How long after you create an agent does it appear? Minutes, hours or the next nightly scan?
- **Write-back.** Can the platform act through the connector (for example block or quarantine), or is it read-only?
- **Permissions needed.** Which roles or scopes does it require, and are they read-only by default?
- **Failure behaviour.** What happens when a token expires? You want an alert, not a silent gap.

Record answers per integration in a simple grid. Our post on [questions to test AI agent discovery coverage](/blog/questions-to-test-ai-agent-discovery-coverage) has a fuller test script, and [acceptance testing for an AI governance platform](/blog/acceptance-testing-for-ai-governance-platform) covers sign-off.

## What open standards should every integration use?

Proprietary connectors are fine for SaaS platforms that have no standard API. Everywhere else, prefer open standards so you are not locked in:

- OIDC and SAML for sign-in; [SCIM 2.0 (RFC 7644)](https://datatracker.ietf.org/doc/html/rfc7644) for users and groups
- OpenAI-compatible endpoints for model traffic
- MCP for tools
- OpenTelemetry for traces and metrics
- REST and webhooks for everything else

## How Opencontroller does this

Opencontroller auto-discovers agents, models and API keys across AWS, Azure and Google Cloud. It has connectors to Salesforce, SAP, Databricks and HubSpot, and to Claude Enterprise and OpenAI Enterprise workspaces, and it discovers and manages skills in Anthropic Enterprise. It supports OIDC SSO, and uses open standards throughout: OpenAI-compatible endpoints, MCP, OpenTelemetry, OIDC/SAML/SCIM, REST and webhooks. Opencontroller Edge covers developer devices and is installed via MDM. Per-agent Block and Quarantine act through the connected platform, so enforcement depth depends on the connector. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**What is the single most important integration?**
Identity. Without users and groups from your identity provider, the platform cannot assign owners, and an inventory without owners cannot be governed.

**Should SIEM integration be required on day one?**
Usually not. Plan it for the first quarter, but require that the platform can export OpenTelemetry traces and webhooks so the SIEM link is straightforward later.

**How do I tell a real integration from a logo?**
Ask for it to run against your test tenant and check read depth, freshness, write-back, required permissions and what happens when credentials expire.

## Key facts

- Day-one integrations for AI agent governance: identity provider (OIDC/SAML plus SCIM), AWS/Azure/Google Cloud, model providers and enterprise workspaces, and SaaS agent platforms.
- First-quarter integrations: code hosts, ITSM/CMDB, SIEM and OpenTelemetry, GRC, developer devices, MCP servers.
- Test each integration for read depth, freshness, write-back, permissions and failure behaviour.
- Opencontroller has connectors to Salesforce, SAP, Databricks, HubSpot, Claude Enterprise and OpenAI Enterprise workspaces, and auto-discovers agents, models and API keys across AWS, Azure and Google Cloud.
- Opencontroller uses OpenAI-compatible endpoints, MCP, OpenTelemetry, OIDC/SAML/SCIM, REST and webhooks.

## Related

- [What questions reveal whether an AI agent discovery tool really covers your estate?](https://opencontroller.lyzr.com/blog/questions-to-test-ai-agent-discovery-coverage.md)
- [What should acceptance testing cover for an AI agent governance platform?](https://opencontroller.lyzr.com/blog/acceptance-testing-for-ai-governance-platform.md)
- [What should an RFP for AI agent governance include?](https://opencontroller.lyzr.com/blog/ai-agent-governance-rfp-requirements.md)
- [More on Buying and comparison](https://opencontroller.lyzr.com/blog/cluster/buying-and-comparison)
