Carbide Analyst connects your compliance program to the tools your team already uses. Ask questions from inside the platform or through any LLM you already work in — the answer comes from your live program either way.
Compliance intelligence with your advisory team in the loop
Carbide Analyst works where you do. Query and take action in your compliance program through any LLM you already use, or work directly inside the Carbide platform.
How it works
Turn your LLM into a compliance engine
The advisory team
AI for speed. Advisors for judgment.
Most compliance AI tools automate tasks and leave the hard calls to you. Carbide Analyst is different because Carbide’s way includes a credentialed advisor who works inside the same platform.
Queries your program, drafts policies, assigns tasks, and updates control statuses. Work that took an hour takes minutes.
Reviews AI-assisted work before it reaches an auditor and signs off on anything that affects your audit outcome. Left unchecked, AI can get things wrong like recommending board minutes as audit evidence when a screen share is the safer call.
Speed without accountability is a liability in compliance. Carbide Analyst and your advisor work inside the same platform, on the same program, so you get both.
What your team can do with Carbide Analyst today
Carbide Analyst works from inside the Carbide platform or through any LLM you already use.
A compliance program that knows when to stop and ask
Most compliance tools automate tasks and leave interpretation to you. Carbide’s architecture is deliberately different: the advisory team works inside the same platform the AI uses, so judgment is built into the process, not bolted on afterward.
What teams ask before getting started
Yes. The Security Assistant is being removed and replaced entirely by Carbide Analyst. It handles everything the Assistant did, plus it can take action inside the platform — updating control statuses, routing evidence, and initiating tasks through natural language — and connects to your LLM of choice via MCP.
A token is what connects your LLM to Carbide. It authorizes your chosen assistant — Claude, ChatGPT, or another MCP-compatible tool — to read from or write to your Carbide account. You create the token under MCP Access in your platform settings, define exactly which tools it can reach, and paste it into your LLM connector. No token means no connection; each token only accesses what you explicitly grant it.
Go to MCP Access in your Carbide settings and revoke or delete the token. This immediately stops Carbide Analyst from interacting with your account through that connection.
Carbide Analyst operates within a defined permission suite that you configure. Access is token-based, granular by tool, and revocable at any time. Full documentation on our AI safety process is available here.
Carbide Analyst is available across all Carbide tiers. Your advisor enables it for your account, contact us to get it turned on.
Yes, and this is intentional. When you create a token, you select exactly which tools it has access to — policies, risk assessments, audit requirements, evidence, or others. You also choose between read-only and read-write access per tool. A read-only token lets you query your program without risking any changes to it. You can run separate tokens for different use cases and revoke any of them instantly from the platform
Yes. The in-platform assistant is context-aware — it knows which page you’re on and which policy or control you have open, so it can provide relevant guidance without you needing to re-explain your situation. When working through your LLM via MCP, it queries your live program data directly rather than relying on a static snapshot.