Intermediate

Task Brain & Automation

Task Brain is OpenClaw's persistent task management layer, introduced in the 2026.3.31 release. Before it, OpenClaw could only handle one-shot requests — you ask, it responds. Task Brain adds the ability to schedule recurring tasks, run background processes, manage parallel subagents, and maintain a durable task history — all stored locally in SQLite.

What Task Brain Manages

SQLite Ledger

Pending tasks

In-progress tasks

Completed tasks

Failed tasks

Schedulers

Cron jobs (recurring)

One-off future tasks

Event-triggered tasks

Agents

Primary agent

Subagent 1

Subagent 2

Subagent N

Task Brain manages all task state, scheduling, and parallel agents

Cron Jobs

Create recurring tasks with natural language — OpenClaw converts them to cron schedules:

“every hour check my email and notify me of urgent messages”

→ 0 * * * *

“every morning at 8am summarize yesterday's GitHub activity”

→ 0 8 * * *

“every weekday at 5pm send me a summary of what I worked on”

→ 0 17 * * 1-5

“every Sunday back up my Documents folder to external drive”

→ 0 2 * * 0

Manage cron jobs with openclaw tasks list to see all scheduled tasks, openclaw tasks pause <id> to suspend one, and openclaw tasks delete <id> to remove it. Or ask the agent via chat: “show me all my scheduled tasks.”

Background Tasks

Some tasks take longer than a single response cycle — compiling a large codebase, processing a big file, running a long web scrape. Task Brain handles these as background tasks: the agent starts the work, tells you it's running, and notifies you via your messaging channel when it's done. You can check status at any time:

You: “process all the PDFs in my Downloads folder and extract key points”
OpenClaw: “Started background task #42. I'll notify you when done (estimated 4 min).”
... 4 minutes later ...
OpenClaw: “Task #42 complete. Processed 23 PDFs. Summary saved to ~/Documents/pdf-summary.md”

Parallel Subagents

For tasks that can be parallelized, Task Brain spawns subagents — each running its own LLM call against a sub-task. The primary agent coordinates:

You: “check the status of all 5 servers in my homelab and give me a summary”
Task Brain spawns 5 subagents, one per server, checking in parallel
Primary agent collects results and sends a unified summary

Subagents use the same model and skills as the primary agent. Each subagent is tracked in the SQLite ledger with its own task ID, so you can audit what ran.

Task History

Every task — one-shot, background, cron, subagent — is written to the SQLite ledger with its input, output, timestamps, and status. This gives you a full audit trail of everything the agent has ever done on your machine. Query it directly with sqlite3 or via the agent: “what tasks did you run yesterday?”

Checklist: Do You Understand This?

Task Brain (2026.3.31) added persistent task management — before it, OpenClaw was one-shot only
Cron jobs: create recurring tasks with natural language; managed via openclaw tasks or chat
Background tasks: long-running work runs detached; you get notified via messaging channel on completion
Subagents: Task Brain parallelizes work by spawning multiple agents on sub-tasks simultaneously
Full audit trail in SQLite — every task is logged with input, output, and timestamps