Skip to content

SplitNest 0.5.0

Latest

Choose a tag to compare

@github-actions github-actions released this 18 Nov 17:18
· 1 commit to main since this release

See what each agent session costs: context, tokens, working time and API-equivalent cost for Claude Code, Codex and Gemini CLI, right in the pane.

  • Agent session usage: a tab where Claude Code, Codex or Gemini CLI runs shows its session in the pane header with the agent's icon: context use, tokens and API-equivalent cost (ctx 21% · 312K tokens · $0.30); click it for each model's tokens (input, cache reads and writes, output), the context against the model's window, when the session started, the agent's working time (Claude Code, Codex), and what the cost means. The tab's and the workspace's tooltips carry the totals. The numbers come from the agent's own session file on this PC, for the conversation SplitNest already finds for Resume; SplitNest keeps only numbers, model names and times from it, never the conversation, and sends nothing anywhere. The cost uses the providers' public API prices shipped with SplitNest (Anthropic, OpenAI, Google; as of 2026-11-18): it is what the tokens would cost on the API, not what a subscription charges, and a model without a known price shows "unknown". When the agent exits, its last numbers stay, marked as ended. Settings → Workspaces → Show agent session usage and Show API-equivalent cost (both on by default) turn them off.

Download

  • Installer: SplitNest-win32-x64-Setup.exe (per-user, no administrator rights, updates itself; see the privacy policy for the update check).
  • Portable: SplitNest-win32-x64-portable.zip (no installer, never checks for updates).

Verify

  • Compare SHA-256 hashes with SHA256SUMS.txt (how). BINARIES.md lists every Windows binary, ours and upstream.
  • These files are built by GitHub Actions from tag v0.5.0 of this repository, with a build provenance attestation: gh attestation verify <file> -R chafian/splitnest.
  • This release is not code signed yet; see the code signing policy.