Platform Integration · Model Context Protocol
Connect Claude to your entire company.
ATLAS exposes your strategies, SaaS spend, security posture, board records, and technology portfolio to your AI assistant via a secure MCP connection. No copy-pasting context. Your AI reads the same live data you do — scoped strictly to your account permissions.
Live context, real answers
Ask questions your AI couldn't answer before
Your AI assistant reads your live ATLAS data across every module it has access to — not a static export, not a manual paste. These examples use real tool calls against the same data structure your organization uses.
"What SaaS contracts renew in the next 60 days, and does any of it map to systems we are planning to replace?"
Upcoming renewals (60 days)
BillForge $14,500/yr — renews in 20 days · Auto-renew ON
EA: maps to "Core Billing" capability. Active migration to Stripe in flight.
Security: 2 active alerts on file.
→ Recommend: cancel renewal, expedite migration.
HubSpot $4,200/mo — renews in 45 days · Auto-renew OFF
EA: maps to "Customer Engagement." No active migration.
→ Standard renewal. Confirm seat count before signing.
Summary
2 renewals in window. Total exposure: $18,700.
1 cancellation candidate. 1 standard renewal.
Setup
Three steps to a connected AI assistant
No local software to install. The ATLAS MCP server is hosted — add a connector in Claude, sign in, and you're done.
Create your ATLAS organization
Set up your organization and populate it with your data — strategies, tech portfolio, SaaS subscriptions, or whichever modules are relevant. The MCP connection is only as useful as the data behind it.
Add the ATLAS connector in Claude
In Claude, go to Settings → Integrations and add a new connector. Enter the ATLAS MCP URL and a name. Click Connect — an ATLAS login page opens. Sign in and authorize the connection.
Ask questions, delegate work
Your AI assistant can now read your live organizational data with your permissions. Give it a task that spans multiple modules — it will call the relevant tools and synthesize a grounded answer.
Step 2 — connector URL
https://atlas-mcp.arcsense.ca/
Works in Claude.ai (web) and Claude Desktop. Enter any name — e.g. "ATLAS".
Coverage
Every module. One connection.
The ATLAS MCP server exposes tools across all seven modules. Read tools are available at every permission level; write tools require Editor or Owner access.
Systems & Strategy
Read: Strategies, goals, outcomes, initiatives, capabilities, technology portfolio, decision log, insights, watchpoints
Write: Create/update strategies, goals, and initiatives; set application lifecycle; link outcomes to initiatives
AI Governance
Read: AI risk register, use case inventory, tool register, guiding principles, vendor assessments, playbook
Write: Log/update risks; register tools; update review status; add use cases; create vendor assessments
Security & Privacy
Read: Security threats, incident log, vendor risk assessments, security artifacts
Write: Log threats and incidents; update threat status; create vendor assessments; update security artifacts
SaaS Operations
Read: SaaS subscriptions with renewal dates, costs, seat counts, auto-renew flags, and notice periods
Write: Subscription management
Corporate Secretary
Read: Legal entity profile, director/officer register, shareholders, board meetings, resolutions, compliance obligations
Write: Corporate record management
Product Management
Read: Products, features with status, roadmap items, stakeholders
Write: Product and feature management
Trust & security
Designed for access you can explain to a security reviewer
MCP access inherits your existing ATLAS permissions and is auditable at the tool level. Read our full security posture.
Role-based access control
Read only
Read-only. Every tool call returns data. No writes permitted — the restriction is enforced at the server boundary, not just the UI.
Read + Write
Read and write. Can create records, update fields, and transition statuses. Every write is validated against an allowlist of permitted fields.
Full access
Full access. Adds the ability to update organization-level configuration and perform administrative operations.
Immutable audit trail
Every write tool call creates an append-only audit record before the response is returned. If the audit write fails, the mutation is not applied — the log and the data are always consistent.
- Tool name, calling user, and timestamp
- Full before/after snapshot of affected records
- Outcome status — success, denied, or error
- Arguments sanitized and stored for replay detection
- Idempotency key support: safe retries never double-write
Open standard
Not a proprietary integration — an open protocol
MCP is an open standard maintained by Anthropic. ATLAS implements it as a hosted server — the same protocol any MCP-compatible AI client can connect to. You are not locked into a single AI provider, and the integration does not depend on any ATLAS-proprietary plugin.
ATLAS data travels directly from the ATLAS API to your AI client. No local software to install, no tokens to rotate manually.
Supported today
Claude Desktop
Claude.ai (Projects)
Any MCP-compatible client
Protocol
Model Context Protocol (MCP)
Open standard — modelcontextprotocol.io
Transport: HTTP + SSE (hosted at atlas-mcp.arcsense.ca)
Common questions
Frequently asked
Direct answers. If something is not implemented yet, we say so.
What is MCP?
The Model Context Protocol (MCP) is an open standard developed by Anthropic that lets AI assistants connect securely to external data sources. ATLAS implements an MCP server so your AI client can query your live organizational data with your permission.
What can my AI see in ATLAS?
Everything your account has access to across all enabled modules — strategies, goals, initiatives, capabilities, technology portfolio, AI tool register, security threats, SaaS subscriptions, board records, product roadmaps, and more. The AI reads exactly what you would see if you were logged in.
Can my AI change data?
Only if your account role permits it. Owner and Editor roles can authorize write-back; Viewer-role connections are strictly read-only. Every write is validated against a field allowlist and logged to an immutable audit trail with a before/after diff.
How does the audit trail work?
Every write tool records the calling user, tool name, arguments, and the full before/after state of the affected record before the response is returned. If the audit write fails, the mutation is not applied — the log and the data are always consistent.
Which AI clients are supported?
Claude.ai (web) and Claude Desktop today. MCP is an open standard — any MCP-compatible client can connect using the same hosted URL. ATLAS does not require you to use a specific AI model or provider.
Is my data used to train AI models?
ATLAS does not share your organizational data with any AI provider for model training. When you use Claude via MCP, Anthropic's standard API data policies apply — your data is processed to generate responses, not used for training by default.
How is the connection secured?
Authentication uses OAuth 2.0 with PKCE. The access token is scoped to a single user identity and expires on a short schedule. All queries are filtered by organization — one user's token cannot read another organization's data. Database-level row security provides a second enforcement layer.
Get started
Give your AI real organizational context.
Create your ATLAS organization, populate it with your data, and connect your AI client. No credit card required.
What the MCP connection does
Reads your live ATLAS data with your account permissions
Calls only the tools you have access to
Logs every write with a before/after diff
Never shares your data with AI providers for training