Chiaro MCP server

Overview

MCP (Model Context Protocol) is the standard way AI tools connect to outside services. Chiaro's MCP server gives your AI the method and the tools it needs to take you through SOC 2.

• It runs in your own AI. Your AI tool does the work on your computer, with access you already have. Chiaro supplies the method and receives what you approve.
• It covers readiness and the audit. Readiness finds your gaps and guides the fixes. In the audit, our team examines your evidence, and it ends in a signed report.
• It works with what you have. Keep the compliance platform you use, and take your work to any auditor.

Connect

Everything runs through one address:

Add it to your AI tool, then sign in to your Chiaro account in the browser window that opens.

Claude Code

Then run /mcp and sign in.

Codex

Then sign in when the browser opens.

Cursor

Click Install in Cursor, then sign in.

Gemini CLI

Then run /mcp auth chiaro and sign in.

Grok bot

Then follow its sign-in steps.

Chiaro works with AI tools that can read files on your own computer, like these five. A chat app in the browser can't reach your files.

How it works

1. Connect and set your scope. Your AI asks a few questions about your product, your systems, the data you handle and which SOC 2 categories apply.
2. Collect. Your AI gathers evidence from wherever it lives: documents in your folders or cloud drive, settings in your systems through read-only commands you approve, and records in your compliance platform.
3. Readiness. Your AI checks each item against our published checks and walks you through the fixes. Your files stay with you.
4. Audit. When you're ready, you start the audit you've chosen and upload copies of your evidence as one zip. Our team examines it.
5. Report. We issue your SOC 2 report, and you can publish a Trust Center page from your portal.

Tools

Your AI works through about 50 tools. Each does one job, and most only read your SOC 2 work or save it. Five can send an email or notification, or start billing, and each of those says so below. No tool can publish your Trust Center page or sign an agreement. You do those yourself, in your portal.

Setup and scope

get_status

Shows where you are and the next step.

select_product

Chooses which of your products this login works on.

start_engagement

Records your company, product line and team size.

set_product_info

Records your product's name, description and customers.

add_system

Adds a system you use and classifies it.

set_data_types

Records the kinds of data you handle.

set_company_profile

Saves your answers to the scoping questions.

confirm_tsc

Sets the SOC 2 categories in scope.

record_hipaa_intake

Saves your HIPAA scoping answers.

get_scope_summary

Shows your scope for review.

confirm_scope

Confirms your scope and builds or updates your checklist.

get_scoping_guidance

Gives vendor questions and evidence examples.

reclassify_system

Changes a system's classification, with a reason.

update_system_description

Fixes a system's description.

record_subservice_soc2

Records whether you have a vendor's SOC 2 report on file.

confirm_scope_delta

For a Type II, records whether your systems changed.

reconcile_scope_inventory

Compares the systems your AI found with your scope.

record_prior_report

Records a first SOC 2, or the facts of a prior report.

record_complementary_controls

Records what you rely on your customers and cloud provider to do.

record_bypass_routes

Answers four yes or no questions about other ways changes can reach your systems.

Evidence

load_area

Loads one area's questions and checks.

submit_evidence

Records a file by its description, findings and fingerprint. The file stays on your computer.

submit_document

The same, for policies and plans.

submit_answer

Saves a written answer.

submit_note

Saves a side comment in your words.

mark_absent

Records that you don't do something.

mark_gap

Logs a gap. Can notify your teammates.

reclassify_evidence

Fixes a file's type or source label.

get_progress

Shows your collection progress.

get_population_spec

Explains what a record list must contain.

submit_population

For a Type II, sends the records for your audit period.

request_upload_link

Makes a time-limited upload link for someone without portal access.

Readiness and audit

start_readiness_check

Shows your readiness map and the next step.

get_judgment_kit

Gives the criteria for one check.

record_self_assessment

Saves your AI's results for checks. Can notify you when every check passes.

record_area_review

Saves an area's results and adopted drafts.

record_policy_adoption

Records your words adopting a policy.

record_population_place

Records where events get logged.

reset_readiness

Clears results and keeps your evidence.

complete_step

Marks collection complete and gives you instructions to upload your evidence zip.

start_audit

Starts the audit you've chosen. Can charge your card or schedule installments, after showing you the terms.

Team

get_questions

Gets the next question for a topic.

answer_question

Saves your answer to a question.

get_unresolved

Checks whether a topic's questions are complete.

resolve_item

Settles a yes or no item.

delegate_topic

Hands a topic to a teammate. Emails them a key for that topic.

undo_delegate

Takes a topic back and revokes its key.

get_topic_delegation_status

Shows who holds which topic.

Questionnaires and help

answer_questionnaire

Drafts answers to a customer's security questionnaire from the facts on file.

record_questionnaire_answers

Saves answers for reuse.

report_problem

Opens a ticket and emails our team.

The list your AI sees when it connects is always the current one.

Your data

• Work happens on your computer. Your AI runs every command on your own computer, with your approval. Chiaro never gets passwords or keys to your systems, and nothing is installed on them.
• During readiness, your files stay with you. They stay in your own folders. For each one, your AI sends us a record of it: a short description, what it found, a fingerprint we can compare with a later copy, and details like the file's location and how it was collected.
• For the audit, you send copies. Your AI gathers copies of the files into one folder on your computer, and you upload it as a single zip from the Upload page in your portal. Those copies are kept in your evidence vault.
• For a Type II, your AI also sends record lists. These are the records for your audit period, like code changes or access granted, for our audit team to test.
• It's your data. You can request a free copy of your evidence vault while your account is open and for 30 days after it closes.
• Where it lives. Your data is stored in the United States, and the AI providers we use are configured not to train on your content.

The full terms are in our Privacy Policy, Terms of Service (section 9) and Data Processing Addendum.

Security and access

• Sign-in. Connecting opens your browser so you can sign in to Chiaro and approve the connection. Your AI then acts with your own role.
• Roles. Admins can do everything. Contributors work only on the topics handed to them. Read-only members can look without changing anything. Every call is checked against your role, and if that check can't run, the call is refused.
• Keys. Admins can also create access keys for an AI tool, and a teammate can get a key limited to one topic, which expires after 90 days. Keys are stored only as a hash.
• Stopping access. Removing Chiaro from your AI tool stops that connection on that computer. Admins can revoke keys on the AI connection page in your portal. Removing a teammate blocks their sign-in access on their next request; revoke any topic keys they hold as well.
• Your systems. Chiaro never connects to your systems or receives credentials for them. Your AI uses access you already have, read-only where possible, and you approve the commands it runs.
• Protection. Data is encrypted in transit and at rest, each customer's data is isolated, production systems are logged and monitored, and backups run nightly. If an incident affects your data, we aim to tell you within 72 hours of confirming it.
• Subprocessors. The vendors we use are listed in our Data Processing Addendum.

Already on Vanta, Drata or Sprinto? Keep it.

You don't have to leave the compliance platform you use. Your AI can connect to your platform's own MCP server alongside Chiaro's, pull what the platform already holds, and send it to us with its source named.

• Vanta. Vanta's MCP server is open to every Vanta user, with access matching your Vanta role. Vanta says it can read your tests, controls, policies and documents.
• Drata. Drata's MCP server can read your evidence library, monitoring tests, controls and policies, within the scopes you grant and your Drata role.
• Sprinto. Once an admin turns it on, Sprinto's MCP server can read your uploaded evidence, your control checks and their pass or fail history.

Your AI uses what your platform already holds, like policies, documents and uploaded evidence. For settings in your own systems, like your cloud console or code host, it reads them straight from the source, which is the strongest kind of evidence. During readiness, your AI checks your evidence against our published checks. During the audit, our team examines the evidence on Chiaro.

What each platform's server can read comes from its own documentation as of October 2026, and it can change. Your AI only sees what your own login on that platform can see.

Take your work to any auditor

• Readiness doesn't lock you in. It's a one-time fee, and the audit is a separate agreement, so you choose who audits you.
• Your work travels with you. Your readiness evidence stays in your own folders as the original files, ready to hand to any CPA firm.
• Our method is public. It's published on GitHub under CC BY 4.0, so another auditor can read what our checks look for.
• Your vault is yours. Ask for a free copy of your evidence vault anytime while your account is open, or within 30 days after it closes.

FAQ