Bits (github.com/harsh071/bits) — vendor claims and pricing as captured 2026-10-30 (UTC+8)
Repo: created 2026-10-21, last push 2026-10-28 03:44 UTC+8, 14 stars (GitHub API, 2026-10-30). Tested main @ cd9e48a (2026-10-28 03:44 UTC+8).
License: README "## License — MIT". No LICENSE file in the repo; GitHub API reports license = null.
Pricing (checked_on 2026-10-30): open source, free. No hosted plan or account. You bring the model: Anthropic or OpenAI API key (paid by usage with those vendors), or a local model through Ollama / LM Studio (free). Product Hunt listing: Free.
Source: https://github.com/harsh071/bits (README), https://www.producthunt.com/products/bits-4

README claims (verbatim):
- "Always on. The daemon runs as a background service. Routines wake it on a schedule, and work continues when the app is closed."
- [C1] "Crash-safe turns. Every step of a turn is saved before the next one. After a crash, the turn picks up where it stopped, and finished tools never run twice."
- [C2] "Steer, don't babysit. Message it mid-task. Risky actions go through the Sentinel, an independent reviewer that only trusts your own words, and ask you only when needed."
- [C3] "Memory in files. ~/.bits/memory/profile.md, a dated fact log, skills/*/SKILL.md and routines/*.md, all Markdown you can read and edit. Searched with SQLite FTS5."
- "Intelligent UI ..." / "Voice ..." (not tested: daemon-only test)
- "Any model. Anthropic, OpenAI, or local models through Ollama / LM Studio (via AI SDK 7). Connect apps with MCP."
- Requirements: "Requires Node 24+ (24.16+ recommended; uses the built-in SQLite) and pnpm."
- Sentinel source comment (apps/daemon/src/agent/sentinel.ts): "The reviewer never sees tool outputs or web content, so injected instructions can't approve actions."