mcp-atlassian: Copilot coding agent retrieves 0 tools from Docker MCP server
When mcp-atlassian (ghcr.io/sooperset/mcp-atlassian:latest) runs as a Docker stdio MCP server in GitHub Copilot Coding Agent on Linux, the log says "Successfully retrieved 0 tools from client: jira-api" even though the server starts and works locally, and another MCP server in the same config (notion-api) lists its tools fine. How do you get the tools to load?
Originally asked on GitHub: https://github.com/sooperset/mcp-atlassian/issues/541
6 answers
Status: still open in the mcp-atlassian repo as of November 2025. The thread points to an interop problem between some third-party clients (GitHub Copilot Coding Agent, Warp) and servers launched over Docker stdio; no project-side fix is confirmed, but two workarounds were reported.
What fixed it for other users
- One commenter fixed it by updating FastMCP to a newer version — mcp-atlassian then worked with the GitHub coding agent.
- Another user put mcp-atlassian behind the TBXark/mcp-proxy remote MCP proxy and reported that resolved the 0-tool retrieval.
- A Warp user got it working by switching to streamable-http instead of stdio:
docker run -p 9000:9000 ghcr.io/sooperset/mcp-atlassian:latest \ --transport streamable-http --port 9000 --host 0.0.0.0with the client config:
json{ "mcpServers": { "mcp-atlassian": { "url": "http://localhost:9000/mcp" } } }Note that a Copilot user who tried the same streamable-http approach still got 0 tools, so results vary by client.
What did not help
- The original reporter ran the Docker stdio config with JIRA_URL, JIRA_USERNAME, JIRA_API_TOKEN and READ_ONLY_MODE passed as env; Copilot connected but retrieved 0 tools while a notion MCP in the same config worked.
- Another user put the token directly in JIRA_API_TOKEN and enabled ENABLED_TOOLS, MCP_VERBOSE and MCP_VERY_VERBOSE for debugging; Copilot still fetched 0 tools.
- A Warp user confirmed the MCP Inspector lists tools correctly, which suggests the server itself responds and the problem is in how some clients call tools/list.
General advice: reproduce with the MCP Inspector first to confirm whether the issue is client- or server-side, keep READ_ONLY_MODE on, and prefer storing the API token in the Copilot environment over hardcoding it in the MCP config.
Did this work for you?What a GitHub user tried on GitHub (Didn't work)
Ran mcp-atlassian via
docker run -iwith JIRA_URL, JIRA_USERNAME, JIRA_API_TOKEN and READ_ONLY_MODE set in the Copilot MCP config. Copilot connected but logged "Successfully retrieved 0 tools", while a notion MCP in the same config listed its tools.Source: https://github.com/sooperset/mcp-atlassian/issues/541 (opens external site)
Did this work for you?What @villelahdenvuo tried on GitHub (Worked)
Updated FastMCP to a newer version. The commenter reported this fixed the issue and mcp-atlassian then worked with the GitHub coding agent.
Source: https://github.com/sooperset/mcp-atlassian/issues/541#issuecomment-3561711929 (opens external site)
Did this work for you?What @JoolsHattey tried on GitHub (Worked)
Put mcp-atlassian behind the TBXark/mcp-proxy remote MCP proxy. The user reported this fixed the 0-tool retrieval.
Source: https://github.com/sooperset/mcp-atlassian/issues/541#issuecomment-3528135151 (opens external site)
Did this work for you?What @davidstevenrojas tried on GitHub (Worked)
Started the container with
--transport streamable-http --port 9000 --host 0.0.0.0and connected the client tohttp://localhost:9000/mcp. The user called it a workaround but confirmed it worked in Warp.Source: https://github.com/sooperset/mcp-atlassian/issues/541#issuecomment-3270628973 (opens external site)
Did this work for you?What @villelahdenvuo tried on GitHub (Didn't work)
Tried the same streamable-http transport from GitHub Copilot Coding Agent. The server started, but Copilot still reported "Successfully retrieved 0 tools".
Source: https://github.com/sooperset/mcp-atlassian/issues/541#issuecomment-3527743439 (opens external site)
Did this work for you?
Related questions
- mcp-atlassian: 'JiraConfig' object has no attribute 'read_only' when calling create_issue
- mcp-atlassian Docker fails on startup with 'ValueError: I/O operation on closed file' — fixed?
- OAuth 2.0 for Confluence Data Center (on-prem) in mcp-atlassian?
- mcp-atlassian crashes with Pydantic 2.12.0: 'cannot specify both default and default_factory'
- mcp-atlassian Jira search_issues: 'requested API has been removed' error and total -1