A2atlassian

What is a2atlassian?

a2atlassian is an MCP server that gives AI agents access to Jira and Confluence. It provides pre-configured connections, read‑only defaults, and compact TSV output to save tokens. It is built for developers and teams who want to let an agent interact with Atlassian tools without Docker or complex .env files.

How to use a2atlassian?

Install a2atlassian with pip install a2atlassian or uv tool install a2atlassian. Use it as an MCP server by running uvx --from a2atlassian a2atlassian-mcp with optional --register or --scope arguments, or as a CLI with commands like a2atlassian login. Configuration is managed via .mcp.json files and environment variable references.

Key features of a2atlassian

• No Docker required – install with pip or uv tool
• Read‑only by default, write access opt‑in per connection
• Pre‑configured connections registered via --register
• Connection scoping with --scope limits agent visibility
• Compact TSV output saves 30‑60% tokens for list results
• Dynamic tool loading reduces context size (Claude Code)
• Error enrichment with field suggestions and auto‑fixed quirks
• Built‑in rate‑limiting retry with exponential backoff

Use cases of a2atlassian

• An agent fetches the status of an issue and adds a comment using Jira tools
• A developer searches Jira issues with JQL and receives compact TSV results
• A team member retrieves Confluence page content and updates page properties
• An agent manages sprints: creates a sprint, adds issues, and updates its state
• Multiple projects are scoped so one agent only sees its own connections

FAQ from a2atlassian

How does a2atlassian differ from other Atlassian MCP servers?

It requires no Docker, uses pip/uv for installation, has ~35 tools loaded on demand, provides read‑only defaults, supports multi‑project scoping, and outputs compact TSV instead of JSON only.

What tools are available in a2atlassian?

There are tools for Jira (read: issues, comments, sprints, boards, worklogs; write: create, update, delete, transition, add worklog) and Confluence (read: pages, children, search; write: upsert pages, set page properties). A full list is in the README.

Is a2atlassian read‑only by default?

Yes, every connection starts read‑only. Write tools are blocked until the operator explicitly sets --no-read-only when saving the connection.

How are credentials stored and handled?

Connections are saved as TOML files at ~/.config/a2atlassian/connections/ with 0600 permissions. Tokens can be stored as environment variable references (${ATLASSIAN_TOKEN}) and are expanded at runtime. Ephemeral --register connections keep credentials only in memory.

What output format does a2atlassian use?

List responses (e.g., search results) use a compact TSV format (header + tab‑separated values) that saves 30‑60% tokens compared to JSON. Single entities (e.g., get issue) return standard JSON with metadata.