Integrations · Business systems
Jira and Confluence agents on VegaDūta
VegaDūta connects to Jira and Confluence through Atlassian's own Remote MCP server, and the connection is genuinely one click: you pick Atlassian from the connector catalog in the console, approve the consent screen on Atlassian's side, and come back to a working connection. There is no API token to copy, and — unlike most OAuth integrations on this platform — no Atlassian app for you to register first, because VegaDūta registers a single platform-wide OAuth client via Dynamic Client Registration and each tenant authorizes against it.
Once connected, whatever Atlassian's server exposes becomes tools your agents can call: look up an issue mid-conversation, read a Confluence page as context, file a ticket as a workflow step. The connection is per-tenant, and you can disconnect it at any time from the same screen.
One click, no token, no app registration
Most business-system integrations start with a scavenger hunt: create an app, copy a client secret, paste an API token. This one does not. VegaDūta holds one OAuth client registered with Atlassian through Dynamic Client Registration, and your tenant authorizes against it with Authorization Code + PKCE. You click Connect, consent on Atlassian's screen, and you are done.
- OAuth 2.1 Authorization Code + PKCE against Atlassian's official Remote MCP server
- No per-tenant Atlassian app to register and no API token to store
- The OAuth state is AES-256-GCM encrypted, so a tampered redirect simply fails to decrypt
- Disconnect from the same screen — the connection is per-tenant and revocable
No long-lived token sitting in a database
This connector deliberately does not store a static authorization header the way a hand-registered MCP server does. The access token is minted fresh on each real connection attempt from the tenant's stored grant, which is encrypted at rest. That keeps the blast radius of a database read smaller than a plaintext-token integration, and means an expired token refreshes on the platform's schedule rather than failing in the middle of a customer conversation.
What an Atlassian-connected agent can do
The tool surface is whatever Atlassian's Remote MCP server publishes for your account — typically searching and reading Jira issues and Confluence pages, and creating or updating them. Your agents call those tools inside the same conversations and workflows as everything else, so a WhatsApp message can become a Jira ticket, or a workflow step can post a Confluence update after a run completes.
Governed like every other connector
Being first-party does not exempt it from the platform's controls. Tool calls run under the same guardrails (PII, toxicity, jailbreak detection), human approval gates can require a person to sign off before an agent writes to your tracker, and every decision lands in the tenant-scoped audit log. Which agents get Atlassian tools at all is your choice — capabilities are per-agent, not global.
Frequently asked questions
How do I connect Jira and Confluence to a VegaDūta AI agent?
Open the connector catalog in the console, choose Atlassian, and approve the consent screen. VegaDūta uses OAuth 2.1 (Authorization Code + PKCE) against Atlassian's official Remote MCP server, so there is no API token to generate and no Atlassian app for you to register beforehand.
Does VegaDūta store my Atlassian API token?
No — there is no API token in this flow at all. The tenant's OAuth grant is encrypted at rest and an access token is minted fresh for each connection attempt rather than being stored as a static authorization header.
What can an agent do in Jira or Confluence?
Whatever Atlassian's Remote MCP server exposes for your account — typically searching and reading issues and pages, and creating or updating them. Those become callable agent tools, usable mid-conversation on any channel or as a workflow step.
Can I stop an agent from writing to Jira without my say-so?
Yes. Human approval gates can require a person to approve a tool call before it executes, guardrails run on the traffic either way, and the decision is audit-logged. You also control which agents have the Atlassian tools enabled at all.
Can I disconnect Atlassian later?
Yes. The connection is per-tenant and revocable from the same screen you connected it on; disconnecting removes the stored grant so agents lose the tools immediately.
See it working in two minutes
The sandbox provisions a real tenant — describe an agent in one sentence and test it, no account, no card. Or browse ~90 industry workflow recipes to see what teams build.
