spawn/daytona
A 9b6e860bba
fix: hetzner destroy_server false failure and daytona silent error swallow (#1622)
Hetzner: grep '"error"' matched '"error": null' in successful DELETE
responses, causing every server deletion to log "Failed to destroy
server" and return 1 even when the server was actually destroyed.
Fix: use jq -e '.action' (consistent with create_server's approach)
which is only present on success.

Daytona: >/dev/null 2>&1 || true suppressed all errors and always
logged "Sandbox destroyed" regardless of outcome. Fix: capture response,
check exit code and error JSON fields, log billing warning on failure.

Same class of bug as AWS (#1606) and GCP/DigitalOcean (#1615).

Agent: team-lead

Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-02-21 18:12:24 -05:00
..
lib fix: hetzner destroy_server false failure and daytona silent error swallow (#1622) 2026-02-21 18:12:24 -05:00
claude.sh fix: use ~/.spawnrc for env vars instead of inlining into .bashrc (#1362) 2026-02-16 17:05:17 -08:00
codex.sh fix: unpin Codex version, restore wire_api=responses (#1608) 2026-02-21 20:12:23 +00:00
kilocode.sh fix: source .spawnrc directly in agent launch commands for reliable env loading (#1546) 2026-02-20 23:37:03 -05:00
openclaw.sh fix: source .spawnrc directly in agent launch commands for reliable env loading (#1546) 2026-02-20 23:37:03 -05:00
opencode.sh fix: source .spawnrc directly in agent launch commands for reliable env loading (#1546) 2026-02-20 23:37:03 -05:00
README.md fix: sync cloud READMEs with current agent list (#1486) 2026-02-19 17:47:57 -05:00
zeroclaw.sh feat: reorder agents and remove NanoClaw (#1477) 2026-02-19 11:39:03 -08:00

Daytona

Daytona sandboxed environments for AI code execution. Daytona

Sub-90ms sandbox creation. True SSH support via daytona ssh. Requires DAYTONA_API_KEY from https://app.daytona.io.

Agents

Claude Code

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/claude.sh)

OpenClaw

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/openclaw.sh)

ZeroClaw

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/zeroclaw.sh)

Codex CLI

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/codex.sh)

OpenCode

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/opencode.sh)

Kilo Code

bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/kilocode.sh)

Non-Interactive Mode

DAYTONA_SANDBOX_NAME=dev-mk1 \
DAYTONA_API_KEY=your-api-key \
OPENROUTER_API_KEY=sk-or-v1-xxxxx \
  bash <(curl -fsSL https://openrouter.ai/labs/spawn/daytona/claude.sh)

Environment Variables

Variable Description Default
DAYTONA_API_KEY Daytona API key (prompted)
DAYTONA_SANDBOX_NAME Sandbox name (prompted)
DAYTONA_CLASS Sandbox class (e.g. small, medium, large) small
DAYTONA_CPU Number of vCPUs (overrides --class) (unset)
DAYTONA_MEMORY Memory in MB (overrides --class) (unset)
DAYTONA_DISK Disk size in GB (overrides --class) (unset)
OPENROUTER_API_KEY OpenRouter API key (OAuth or prompted)

Note: Daytona rejects explicit --cpu/--memory/--disk flags when using snapshots. Use DAYTONA_CLASS instead. If explicit resource flags fail due to snapshot conflict, spawn automatically retries with --class small.