- We have an internal AI application
- We want to connect it with Jira Cloud via MCP
- Authentication should be OAuth-based (per user login)
MCP Server URL used:https://mcp.atlassian.com/v1/mcp and https://mcp.atlassian.com/v1/mcp/authv2
- We also have an OAuth 2.0 (3LO) app created in Atlassian Developer Console
- Validation failed Authentication Failed Missing PKCE verifier in state
-
- Looks like MCP requires OAuth 2.1 with PKCE flow
- However, our platform validation seems to be happening server-to-server without a browser redirect
- We are unsure:
- Whether MCP supports integration from custom backend/UI tools
- Whether we should use our own 3LO OAuth app or MCP-managed OAuth
- How PKCE state is expected to be handled in such integrations
-
- Can Atlassian MCP Server be used with custom AI/backend applications, or is it only supported with MCP-native clients (like Claude, VS Code, etc.)?
- Should we use our own OAuth 2.0 (3LO) app, or is MCP expected to manage OAuth internally?
- How should the PKCE flow (code_verifier/state) be handled when the client is not a browser-based MCP client?
- Is there any supported pattern or reference implementation for integrating MCP with custom applications?