Atlas developer platform

A first-class API, and it works with Claude.

Ship on an OpenAPI 3.1 REST API with HMAC-signed webhooks, OAuth 2.0 apps, and idempotency keys, or skip the SDK entirely and add Atlas as an MCP server so Claude and any agentic client can read and write your workspace.

Read the API docs MCP docs

OpenAPI 3.1
Signed webhooks
OAuth 2.0
Works with Claude via MCP

Create a taskREST + idempotency

curl -X POST https://api-atlas.wrxstack.com/v1/tasks \
  -H "Authorization: Bearer atlas_pat_live_..." \
  -H "Idempotency-Key: 7f3a91c2-1b8e-4d3f-9c6a" \
  -H "Content-Type: application/json" \
  -d '{ "title": "Review contract", "priority": "high" }'

Or add Atlas as an MCP server in Claude and skip the SDK entirely.

How it works

Two ways to build, one workspace.

Use the REST API for classic integrations, or connect an agent over MCP. Both are scoped per tenant and written to the audit log.

01
Authorize
Create a scoped personal access token or an OAuth 2.0 app with granular scopes.
02
Call the API
OpenAPI 3.1 endpoints with typed clients for TypeScript, Python, and Go.
03
Stay consistent
Idempotency keys on every write make retries safe and duplicates impossible.
04
Subscribe to events
HMAC SHA-256 signed webhooks with replay protection push changes to your app.
05
Or connect over MCP
Add Atlas as an MCP server in Claude Desktop or Claude Code, no SDK required.
06
Audit everything
Every API call and MCP tool call lands in the tamper-evident audit log.

What you get

The integration layer is a product, not an afterthought.

OpenAPI 3.1
A documented spec with typed clients for TypeScript, Python, and Go.
Signed webhooks
HMAC SHA-256 signatures, delivery IDs, and replay protection on every event.
OAuth 2.0 apps
Scoped tokens, per-app permissions, and rotation, for building on behalf of tenants.
Built-in MCP server
Read and write tasks, projects, notes, files, and calendar from any MCP client.
Scoped and audited
Per-tenant, tool-level scopes with an audit entry on every call.
Rate limits + pagination
Per-tenant rate limits, cursor pagination, filtering, and sorting out of the box.

Explore more

Everything you need to build.

FAQ

Questions, answered.

Does Atlas really work with Claude?: Yes. Atlas ships a built-in Model Context Protocol server, so you can add it to Claude Desktop or Claude Code and let Claude read and write tasks, projects, notes, files, and calendar, scoped to your tenant and logged in the audit trail.
What does the REST API cover?: The API is documented with OpenAPI 3.1 and covers the core workspace objects, with typed clients for TypeScript, Python, and Go, HMAC-signed webhooks, OAuth 2.0 apps with scoped tokens, idempotency keys on every write, and per-tenant rate limits.
How are webhooks secured?: Every webhook is signed with HMAC SHA-256 and carries a delivery ID and timestamp, with replay protection, so your endpoint can verify authenticity and reject duplicates.
Is the MCP server scoped and safe?: Yes. MCP tokens are per-tenant with tool-level scopes (read, write, admin), responses stream over the MCP transport, and every tool call is written to the tamper-evident audit log.

Ready when you are

Build on Atlas, or let Claude do it.

Start free, grab a scoped token, and ship an integration over REST, or add Atlas to Claude over MCP in minutes.

Read the API docs Start free

A first-class API, and it works with Claude.

OpenAPI 3.1

Signed webhooks

OAuth 2.0

Works with Claude via MCP

curl -X POST https://api-atlas.wrxstack.com/v1/tasks \ -H "Authorization: Bearer atlas_pat_live_..." \ -H "Idempotency-Key: 7f3a91c2-1b8e-4d3f-9c6a" \ -H "Content-Type: application/json" \ -d '{ "title": "Review contract", "priority": "high" }'

Questions, answered.

Does Atlas really work with Claude?

Yes. Atlas ships a built-in Model Context Protocol server, so you can add it to Claude Desktop or Claude Code and let Claude read and write tasks, projects, notes, files, and calendar, scoped to your tenant and logged in the audit trail.

What does the REST API cover?

The API is documented with OpenAPI 3.1 and covers the core workspace objects, with typed clients for TypeScript, Python, and Go, HMAC-signed webhooks, OAuth 2.0 apps with scoped tokens, idempotency keys on every write, and per-tenant rate limits.

How are webhooks secured?

Every webhook is signed with HMAC SHA-256 and carries a delivery ID and timestamp, with replay protection, so your endpoint can verify authenticity and reject duplicates.

Is the MCP server scoped and safe?

Yes. MCP tokens are per-tenant with tool-level scopes (read, write, admin), responses stream over the MCP transport, and every tool call is written to the tamper-evident audit log.

A first-class API, and it works with Claude.

Two ways to build, one workspace.

Authorize

Call the API

Stay consistent

Subscribe to events

Or connect over MCP

Audit everything

The integration layer is a product, not an afterthought.

OpenAPI 3.1

Signed webhooks

OAuth 2.0 apps

Built-in MCP server

Scoped and audited

Rate limits + pagination

Everything you need to build.

Questions, answered.

Build on Atlas, or let Claude do it.

A first-class API, and it works with Claude.

Two ways to build, one workspace.

Authorize

Call the API

Stay consistent

Subscribe to events

Or connect over MCP

Audit everything

The integration layer is a product, not an afterthought.

OpenAPI 3.1

Signed webhooks

OAuth 2.0 apps

Built-in MCP server

Scoped and audited

Rate limits + pagination

Everything you need to build.

Questions, answered.

Build on Atlas, or let Claude do it.