Basecamp becomes agent accessible

· ai ai-agents · Source ↗

TLDR

  • 37signals has shipped MCP-based agent access for Basecamp, enabling agents to read, write, and manage projects programmatically.

Key Takeaways

  • Agents can now create to-do lists, post messages, upload files, chat, and arrange project schedules in Basecamp – full feature parity with human users.
  • 37signals experimented with embedded AI features in Basecamp and Fizzy for 18 months but did not ship any, citing poor quality and user acceptance.
  • MCP enables cross-tool agent workflows: example given is Sentry bug data piped into a Basecamp message via agent, without hand-written integration code.
  • Basecamp is first; Fizzy and HEY are next. 37signals is also shipping revamped APIs and CLIs across products.
  • DHH argues agent adoption will be driven by mainstream interfaces (ChatGPT, Gemini, Grok) embedding agents, not power users running Claude Code or Gemini CLI directly.

Why It Matters

  • The same workflows that previously required custom API integrations and data engineers are now executable by general-purpose agents with no bespoke code.
  • A vanishingly small share of Basecamp customers have ever used its API directly; agent-driven access is expected to reach a much larger share much faster.
  • 37signals is treating agent accessibility as infrastructure-level, not a feature – revamped APIs and CLIs are part of the same push.

David Heinemeier Hansson · 2026-03-25 · Read the original

Related coverage