LONDON, 5 May 2026: SprintFlint today released v1.7. The headline: the public Model Context Protocol server is live again at mcp.sprintflint.com, speaking MCP protocol 2025-06-18 over Streamable HTTP. Drop one block of JSON into Claude Desktop and the agent can list projects, create sprints, file tickets, and post comments without ever leaving the editor.
v1.4 was the content engine. v1.5 was the practitioner engine. v1.6 was the discovery layer. v1.7 is the agent surface: SprintFlint is now natively addressable by every MCP-aware AI tool, alongside the existing REST API.
MCP server live at mcp.sprintflint.com
The relaunch resolves a long-standing routing failure. The MCP app shares the SprintFlint codebase but ships from its own Procfile.mcp, selected at build time via the multi-procfile buildpack. The web process boots mcp.ru instead of the standard Rails router, and Heroku ACM issued a Let's Encrypt certificate for the custom domain. The verified initialize handshake returns sprint-flint-rails v0.0.1 with tools, prompts, resources and logging capabilities advertised.
Connect Claude Desktop in 30 seconds
{
"mcpServers": {
"sprintflint": {
"url": "https://mcp.sprintflint.com",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}Same bearer token as the REST API. Same scopes. No separate auth flow. Cursor, Claude Code, Zed and any other MCP-capable client work the same way.
New long-form posts
- Agile Estimation Techniques Compared; story points vs t-shirt sizing vs ideal hours vs no estimates, with a decision matrix for which to use when.
- Sprints With AI Teammates; a working playbook for treating Claude Code, Cursor and Codex as sprint participants: ticket assignment, definition-of-done, review etiquette, and where the human still owns the call.
Atom feed for the blog
/blog/feed.atom exposes the full blog as a standards-compliant Atom 1.0 feed, auto-discovered by readers via <link rel="alternate"> in the blog index. NetNewsWire, Feedly, Reeder and the AI agents that watch RSS for content discovery now all light up.
Marketing polish; the bits that bug you
- Header on every marketing page. Blog, tools, glossary, templates, migrations, integrations, roadmap and error pages now all carry the site header. No more dead-end navigation.
- 4-column grouped footer. Product, Resources, Company, Legal; replaces the previous single-column wall of links. Two columns on mobile, four on desktop.
- Unified tools CTA. Every public tool page ends with the same signup card, except when embedded.
- Mobile fix on /api Resource model cards. Long inline endpoint paths no longer break the grid on narrow screens.
- Turbo cache fix. Opening the landing-page demo video, navigating away, then hitting back used to leave a transparent modal trapping every click. The modal now closes itself before Turbo snapshots the page.
Why this matters for ICP
SprintFlint's ideal customer is the engineering manager or tech lead whose team already lives inside an AI coding agent. v1.7 closes the loop: their agent can now read and write SprintFlint state directly via MCP, while the marketing surface (header, footer, blog feed) finally behaves like a site rather than a maze. The content compounds. The agents have a stable interface. The pages link back to themselves.
Pricing unchanged
Free up to 300 tickets, forever. No card, no countdown. £5/user/month or £50/user/year if you outgrow the free tier. MCP server, REST API and every free tool, blog post, template and glossary term remain free forever, no signup, no email gate.
Try it
sprintflint.com for the product. mcp.sprintflint.com for the MCP endpoint. /integrations for the per-editor setup guides. /api for the REST surface. Free signup at /magic-link/new.