Tasks
What is a task?
A task is the smallest unit of work a Gobii agent performs. Every action your agent takes — searching the web, reading a page, running a browser automation, sending an email, calling an API — counts as one or more tasks.
Tasks happen behind the scenes. You don't need to manage them directly, but understanding them helps you interpret what your agent is doing and how it consumes credits.
How tasks work
When you ask your agent to do something, it breaks the request into a sequence of tasks. Each task is a discrete tool call: one search, one page scrape, one email send, one chart generation.
Example
You ask: "Look up the weather in Rehoboth Beach, DE this weekend and find appropriate daytime activities for each day."
Your agent plans and executes these tasks:
- Search for the weekend weather forecast for Rehoboth Beach
- Extract and parse the forecast data
- Use the LLM to match activities to the forecast
- Search for venues matching each activity
- Compose and send an email with the results
That's 5 tasks to fulfill one request. Some tasks are fast (milliseconds for an API call), others take longer (browser automation can run for minutes).
Types of tasks
Tasks span the full range of agent capabilities:
| Category | Examples |
|---|---|
| Search & browse | Web searches, page scraping, reading documentation |
| Browser automation | Logging into sites, filling forms, capturing screenshots, clicking through workflows |
| Messaging | Sending emails, SMS, chat messages |
| Data | Running SQL queries, generating charts, creating CSV/PDF exports |
| Files | Reading, writing, and patching files in the agent workspace |
| Code | Executing Python or shell commands in the sandbox |
| Integrations | Calling MCP servers, connected apps (Google Sheets, HubSpot, Discord, etc.) |
| Planning | Updating the agent's plan, requesting human input |
Tasks vs. task credits
Tasks consume task credits from your account. Most tasks cost a fraction of a credit — the base cost per task is 0.150 credits, and most individual tool calls (searches, scrapes, API calls, messages) fall in this range.
The actual cost of a task is:
Intelligence tiers and multipliers
Your agent's intelligence level multiplies the base cost of every task:
| Tier | Multiplier | Cost per task | Best for |
|---|---|---|---|
| Standard (1×) | 1.00 | 0.150 credits | Routine tasks, simple lookups |
| Premium (2×) | 2.00 | 0.300 credits | Research, multi-step workflows |
| Max (3×) | 3.00 | 0.450 credits | Complex analysis, deep research |
| Ultra (8×) | 8.00 | 1.200 credits | Maximum capability |
| Ultra Max (15×) | 15.00 | 2.250 credits | Peak performance |
Higher tiers give your agent access to more capable models and deeper reasoning, but they consume credits faster. You set the intelligence level per agent in the agent settings.
Example: On the Standard tier, a typical task costs ~0.150 credits. On Max (3×), the same task costs ~0.450 credits — still fractional, but proportional to the intelligence boost.
Your plan includes a monthly pool of task credits. See Task Credits for details on budgets, soft limits, and monitoring usage.
Task execution
Agents execute tasks sequentially within each processing cycle. You can watch tasks unfold in the agent's chat timeline — each tool call appears as it happens, with results and any errors.
If a task fails (e.g., a page won't load, an API returns an error), the agent typically retries or adjusts its approach. Persistent failures are reported in the chat so you can intervene.
Agent budgets and limits
Each agent has a daily soft limit on task credits. This prevents runaway consumption — especially important at higher intelligence tiers where each task costs more. The agent will pause when it hits the limit and resume the next day.
You set this limit per agent in the agent settings. See the agent settings guide for details.
Monitoring tasks
Tasks appear in the agent's chat timeline as a collapsed "X actions" button. Click it to expand and see each individual task the agent performed — including search queries, page scrapes, tool calls, and messages — with their results and credit costs.
- Chat timeline: Watch tasks execute in real time as your agent works; the "X actions" button shows task-level detail
- Usage dashboard: View aggregate task credit consumption at
https://gobii.ai/console/usage/ - Agent logs: Each agent's activity history shows completed tasks and their credit costs
Next steps
- Understand task credits — budgets, soft limits, and billing
- Configure agent budgets — set daily limits per agent
- See tasks in action — create an agent and watch it work