OpenClaw Cron Fallback Simulator
Plan and simulate cron-scheduled subagent tasks with model fallback chains. Define schedules, build fallback priorities, configure per-task failure notifications, and simulate model outages — then export your config. Based on 6 OpenClaw PRs merged May 15, 2026.
Why Use the Cron Fallback Simulator
Built around the cron scheduling and model fallback improvements merged into OpenClaw on May 15, 2026.
Cron Schedule Builder
Define cron expressions for subagent tasks with a live human-readable preview. Plan how often your agents run and when.
Model Fallback Chains
Build ordered model fallback chains per task. When the primary model is unavailable, OpenClaw automatically tries the next in line (PR #82302).
Shared Fallback Selection
PR #82328 consolidated how cron tasks share fallback logic. The simulator shows how models resolve across multiple concurrent tasks.
Failure Simulation
Set per-model failure rates and run simulations to see how your fallback chain handles real-world outages over time.
Separate Failure Notifications
PR #82297 lets each task route failure alerts to its own channel instead of flooding a shared notification channel.
Wildcard Model Policy
PR #82282 ensures wildcard model runtime policies apply correctly. Toggle per-task to see how wildcards affect resolution.
Simulate Your Cron Fallback Setup
Add cron tasks, build model fallback chains, tune failure rates, then run the simulator to see how your setup handles outages.
Cron Tasks
Define cron-scheduled subagent tasks with model fallback chains. When the primary model fails, OpenClaw tries the next in the chain (PR #82302). Reorder models to set fallback priority.
Simulation Settings
The simulator fires each task at its cron schedule, tries models in fallback order, and tracks which model handles each execution. Adjust Fail % per model to simulate outages and see how the fallback chain resolves.
Generated Configuration
Add this to your OpenClaw config to set up cron tasks with model fallback chains. Includes shared fallback selection (PR #82328), separate failure notifications (PR #82297), and session context delivery (PR #82241).
# OpenClaw cron subagent fallback config
# Shared fallback selection (PR #82328)
# Honors subagent model fallbacks (PR #82302)
# Separate failure notification delivery (PR #82297)
agents:
cron:
- name: "Daily Summary Agent"
schedule: "*/10 * * * *"
models:
- provider: "OpenRouter"
model: "anthropic/claude-opus-4-6"
- provider: "OpenRouter"
model: "openai/gpt-5.3-codex"
- provider: "OpenRouter"
model: "qwen/qwen3.5-plus"
failureNotification:
mode: "separate" # PR #82297
target: "#alerts-summary"
deliveryContext: session # PR #82241Deploy OpenClaw with DeployClaw
Get automatic cron scheduling, model fallback chains, separate failure notifications, and real-time monitoring out of the box.
Frequently Asked Questions
What changed with cron fallbacks in OpenClaw PRs #82328 and #82302?
PR #82328 refactored how subagent cron tasks share their fallback model selection logic, consolidating it into a shared module. PR #82302 fixed a bug where scheduled agent tasks weren't respecting configured model failover options — so if your primary model was down, the cron job would fail instead of trying the next model in the chain. Together, these PRs make cron fallback behavior consistent and reliable.
How does separate failure notification delivery work (PR #82297)?
Previously, all cron task failure notifications were routed through a single shared channel, which could cause alert fatigue when multiple tasks failed simultaneously. PR #82297 adds per-task notification routing so each cron task can send failure alerts to its own channel — for example, your 'Daily Summary Agent' alerts go to #alerts-summary while your 'Code Review Bot' alerts go to #alerts-reviews.
What is the wildcard model runtime policy (PR #82282)?
The wildcard model runtime policy lets you define catch-all rules for model selection. For example, setting a wildcard policy means any model matching a pattern (like 'anthropic/*') gets the same runtime constraints. PR #82282 fixed a bug where these wildcard policies weren't being applied during agent execution, causing unexpected model behavior in fallback chains.
How does session context delivery resolution work (PR #82241)?
PR #82241 refactored how cron tasks resolve their delivery targets from the session context. Instead of determining the delivery channel at schedule-parse time, the resolution now happens when the task actually fires — meaning if a session's context changes (e.g., a channel goes offline), the cron task adapts to the current state rather than using stale routing information.
What are OpenAI completions token aliases (PR #82278)?
Different providers use different field names for the same concept — for example, OpenAI uses 'max_tokens' while some compatible APIs use 'max_completion_tokens'. PR #82278 ensures OpenClaw maps these aliases correctly in fallback chains, so switching from an OpenAI model to a compatible provider doesn't break due to field naming differences.
How many fallback models should I configure per cron task?
For most deployments, 2-3 fallback models is optimal. Your primary model handles the normal case, a second model from a different provider catches single-provider outages, and an optional third serves as a last resort. Use the simulator to test different failure scenarios — you'll see diminishing returns beyond 3-4 models, and more models add complexity to debugging.