spike-mcp

# spike-mcp [](https://pypi.org/project/spike-mcp/) [](https://pypi.org/project/spike-mcp/) MCP server that connects Claude to Jira and Confluence. Engineers can research spikes, generate technical solutions with Mermaid architecture diagrams, write Confluence docs, and create Jira epic + story breakdowns through natural conversation — no Anthropic API key required. --- ## Features - **Research** — search Confluence and Jira before generating anything - **Design** — produce Mermaid architecture/flow diagrams as part of every spike doc - **Write** — create or update Confluence pages (markdown + Mermaid → Confluence storage format) - **Ticket** — create Jira Epics, Stories with acceptance criteria, and Tasks; supports both next-gen and classic projects - **Zero LLM cost** — Claude itself provides all intelligence; no separate AI API key needed --- ## Installation ```bash pip install spike-mcp # or uv add spike-mcp ``` --- ## Setup ### Step 1 — Create `~/.spike.toml` Create `~/.spike.toml` in your home directory with your Atlassian details: ```bash # macOS / Linux touch ~/.spike.toml ``` ```toml [atlassian] base_url = "https://yourorg.atlassian.net" email = "you@yourorg.com" [confluence] # If your Confluence is on a different domain than Jira, set this: # base_url = "https://yourorg-docs.atlassian.net" space_key = "ENG" parent_page_id = "123456" [jira] project_key = "PLAT" epic_issue_type = "Epic" story_issue_type = "Story" task_issue_type = "Task" default_label = "spike" story_points_field = "customfield_10016" epic_link_field = "customfield_10014" [tickets] story_point_scale = [1, 2, 3, 5, 8, 13] ``` > **Tip:** Config is also discovered automatically in your project root or any parent directory up to the git root — useful when running spike-mcp from a specific repo. The search order is: current directory → git ro…