eazybi-mcp

A read-only Model Context Protocol server that lets LLMs query eazyBI — Reports & Charts for Jira. Plug it into Claude Desktop, Claude Code, Cursor, or any other MCP-compatible client and ask things like:

"Pull the Cycle Time report for Q1 2026 from eazyBI and tell me what changed." "List the dashboards in account 250143 and summarise the velocity ones." "Export report 5259102 as CSV to my Downloads folder."

The server is intentionally read-only — it cannot modify reports, dashboards, or data — so giving it an API token is low-risk.

Why

eazyBI builds incredibly powerful Jira analytics, but its data lives behind an OLAP cube and a UI. If you want an LLM to reason over your team's velocity, cycle time, throughput, or any custom metric you've already modeled in eazyBI, you have two options: re-implement the analysis against the Jira API (slow, brittle, ignores your calculated members), or run the saved report you already have and feed the result into the LLM. This server does the second.

What's stable vs experimental

eazyBI publishes exactly one REST endpoint as part of its supported public API: the Report Results Export API. That is what export_report and get_export_url wrap. eazyBI Support has stated explicitly that there is no REST API for listing or publishing reports / dashboards — not on Cloud, not on Data Center, not on eazyBI.com.

The remaining list_* / get_* tools probe internal Rails JSON routes (/accounts.json, /accounts/{id}/cubes.json, …) that the eazyBI UI itself consumes. They are not part of any documented contract:

• on Atlassian Cloud these routes reject Basic auth — they always return {"supported": false} and you should not rely on them;
• on Data Center / Private eazyBI they often happen to work (Jira's plugin framework gates everything with the same auth) but eazyBI Support recommends against treating them as a stable interface.

So the realistic picture is: export_report is the rock-solid path; everything else is a "good luck" probe of UI internals. The export tool is enough for almost every reporting workflow — if you know account_id and report_id, you can run any saved report end-to-end.

Tools

Tool	Status	What it does
export_report	stable	Run a saved report by ID and return data as JSON / CSV / XLS / PDF / PNG. Supports selected_pages for page-filter overrides and embed_token for publicly shared reports.
get_export_url	stable	Build the canonical export URL for a report — no HTTP request made.
list_accounts	experimental	List eazyBI accounts visible to the authenticated user.
list_reports	experimental	List reports in an account, optional folder filter.
get_report_definition	experimental	Full JSON definition of a report (rows / columns / pages / calculated members).
list_dashboards	experimental	List dashboards in an account.
get_dashboard	experimental	Layout of a dashboard and the IDs of the reports it contains.
list_cubes	experimental	List OLAP cubes available in an account.
get_cube	experimental	Full cube metadata: dimensions, hierarchies, measures.
list_dimensions	experimental	Dimensions of a cube.
list_measures	experimental	Measures of a cube, optional folder filter.
list_hierarchies	experimental	Hierarchies of a single dimension.

Quickstart

Pick one of the install options below, then hand the binary to your MCP client.

Install with uv (recommended)

brew install uv     # one-time
uv tool install --from git+https://github.com/ultimate-guitar/eazybi-mcp eazybi-mcp

Install with pipx

pipx install git+https://github.com/ultimate-guitar/eazybi-mcp

From source

git clone https://github.com/ultimate-guitar/eazybi-mcp
cd eazybi-mcp
python3 -m venv .venv
source .venv/bin/activate
pip install -e .

Configure

Copy .env.example to .env (or set the variables in your MCP client's env block) and fill in:

EAZYBI_BASE_URL=https://aod.eazybi.com         # eazyBI for Jira Cloud
[email protected]
EAZYBI_API_TOKEN=your-atlassian-api-token      # https://id.atlassian.com/manage-profile/security/api-tokens
EAZYBI_DEFAULT_ACCOUNT_ID=12345                # optional, looked up from the URL after /accounts/

Deployment	EAZYBI_BASE_URL	Auth
eazyBI for Jira Cloud	https://aod.eazybi.com	Atlassian email + API token
eazyBI for Jira / Confluence Data Center	https://your-jira.example.com/plugins/servlet/eazybi	Jira/Confluence username + password (or PAT)
eazybi.com (SaaS)	https://eazybi.com	eazyBI account email + password

Cloud users: the URL inside the eazyBI iframe in Jira looks like https://aod.eazybi.com/eazy/accounts/.../.... The /eazy segment is for UI routes only — set EAZYBI_BASE_URL without it. The export endpoint lives at the bare host.

Public reports: if a report is shared via "Embed report → Public access token", pass embed_token="..." to export_report and skip Basic auth.

Where to find account_id and report_id

Open a report in eazyBI; the URL looks like …/accounts/250143/cube_reports/5259102 (or …/reports/5259102-foo). The first number is account_id, the second is report_id. The "Embed report" action in the eazyBI toolbar surfaces both as well.

Wire it into a client

Claude Desktop

Edit claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "eazybi": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/ultimate-guitar/eazybi-mcp", "eazybi-mcp"],
      "env": {
        "EAZYBI_BASE_URL": "https://aod.eazybi.com",
        "EAZYBI_USER": "[email protected]",
        "EAZYBI_API_TOKEN": "your-atlassian-api-token",
        "EAZYBI_DEFAULT_ACCOUNT_ID": "12345"
      }
    }
  }
}

For a globally-installed binary use "command": "eazybi-mcp", "args": [] instead.

After editing, fully quit Claude Desktop (Cmd+Q on macOS) and reopen it — closing the window is not enough.

Claude Code

claude mcp add eazybi -- eazybi-mcp \
  --env EAZYBI_BASE_URL=https://aod.eazybi.com \
  --env [email protected] \
  --env EAZYBI_API_TOKEN=your-atlassian-api-token

Cursor / Continue / any MCP client

Point the client at the eazybi-mcp binary; pass the env vars in whatever way your client supports.

Sample prompts

Once wired up:

• "What's the P85 cycle time for Q1 2026? Pull report 5259102 from eazyBI."
• "List all reports in the 'Velocity' folder of my eazyBI account."
• "Export report 5259102 as CSV and save to ~/Downloads/cycle-time.csv."
• "Compare the last three quarters of cycle time and tell me if the trend is up or down."

The LLM picks tools automatically. For Cloud users only export_report is reliable — that's enough for most reporting tasks.

Driving the client directly (for tests / scripts)

import asyncio
from eazybi_mcp.client import EazyBIClient, EazyBIConfig

async def main():
    client = EazyBIClient(EazyBIConfig.from_env())
    payload, ctype = await client.export_report(
        account_id=250143,
        report_id=5259102,
        fmt="json",
    )
    print(ctype, len(payload))
    await client.aclose()

asyncio.run(main())

Limitations

• eazyBI does not expose a public API to fetch dashboard data as a single payload. Use get_dashboard (Data Center / Private only) to discover the constituent report IDs, then call export_report for each.
• Arbitrary MDX execution is intentionally not wired up — eazyBI does not document a public endpoint for it. If your private deployment exposes one, extend EazyBIClient accordingly.
• Large JSON exports are truncated by the max_chars parameter. Pass save_to=/path/to/file.json to write the full payload to disk.
• This is an unofficial integration. eazyBI may change the experimental routes at any time; the stable export endpoint is the long-term contract.

Contributing

Issues and PRs welcome. Please don't commit your .env or any token — the project's .gitignore excludes .env precisely for that reason. To add a new tool:

1. Add the HTTP call to EazyBIClient in src/eazybi_mcp/client.py.
2. Wrap it as a @mcp.tool() in src/eazybi_mcp/server.py.
3. Mark it [stable] only if it hits a documented public endpoint; otherwise [experimental] and route it through _safe_experimental so 401/404 responses degrade gracefully.

License

MIT