Optimizing Credit Usage

Credit usage is one of the most common concerns for Gobii users. This guide covers practical strategies to get the most value from every credit — from choosing the right intelligence tier to designing efficient agent workflows.

Understanding the Cost Model

Every task performed by a Gobii agent costs credits. The cost depends on two factors:

Actual Cost = Base Task Cost × Intelligence Multiplier

• Base task cost: 0.150 credits per task (the default across all plans)
• Intelligence multiplier: Determined by the agent's intelligence tier setting

Per-Task Costs by Intelligence Tier

Tier	Multiplier	Per-Task Cost	Best For
Standard	1×	0.15 credits	Simple lookups, monitoring, routine data collection, scheduled scrapes
Premium	2×	0.30 credits	Research synthesis, multi-step analysis, summarization
Max	3×	0.45 credits	Complex reasoning, code review, detailed report generation
Ultra	8×	1.20 credits	Deep analysis, large-context coordination, advanced problem-solving
Ultra Max	15×	2.25 credits	Maximum reasoning power, complex multi-agent orchestration

The math is simple: a single Ultra Max task (2.25 credits) costs 15× more than a Standard task (0.15 credits). If your agent runs 100 scheduled tasks per day at Ultra Max, that's 225 credits — versus 15 credits at Standard.

Strategy 1: Choose the Right Intelligence Tier

The single biggest lever for credit optimization is the intelligence tier. Most agents don't need the highest tier for every task.

When to Use Standard (1×)

Standard handles straightforward, well-defined tasks reliably. Use it for:

• Scheduled monitoring — Checking prices, news headlines, or status pages
• Data collection — Scraping structured data, pulling API results
• Simple formatting — Converting data between formats, basic transformations
• Webhook handling — Processing incoming webhook payloads

When to Use Premium (2×)

Premium adds better reasoning for multi-step tasks. Use it for:

• Research synthesis — Reading multiple sources and summarizing findings
• Email drafting — Composing context-aware responses
• Conditional workflows — Branching logic based on results

When to Use Max+ (3× and above)

Save Max, Ultra, and Ultra Max for tasks that genuinely need deep reasoning:

• Code review and debugging
• Complex analysis with many variables
• Large-context document processing
• Multi-agent coordination

Tip: Start agents at Standard or Premium. Only increase the tier if you see the agent making reasoning errors on its current tasks. You can always bump it up later.

Strategy 2: Set Effective Daily Budgets

Every agent has a configurable daily soft limit for credit usage. When the agent approaches this limit, Gobii aims to stay within it. A hard limit (derived from the soft limit — 2× by default) stops the agent for the day once reached.

Budgeting Guidelines

Agent Type	Suggested Daily Limit	Reasoning
Simple monitor (Standard, once/hour)	5–10 credits	~33–66 tasks/day covers hourly checks
Research assistant (Premium, scheduled briefings)	10–20 credits	Allows a few research sessions per day
Multi-agent coordinator	20–50+ credits	Coordination tasks are deeper; adjust based on observed usage

Start Conservative

1. Set a low initial budget (e.g., 10 credits/day)
2. Monitor actual usage on the Usage dashboard for a few days
3. Increase only if the agent is consistently hitting its limit while producing valuable work
4. If the agent rarely hits its limit, consider lowering it further

Plan-Level Defaults

Each plan has a default daily credit target, but you can adjust it per agent. The slider lets you set a soft limit with a hard limit (typically a multiple of the soft limit) as a backstop.

Strategy 3: Design Credit-Efficient Workflows

Batch Work Instead of Spreading It

An agent that runs a single comprehensive task once uses fewer credits than an agent that runs many small tasks. For example:

• ❌ Inefficient: An agent that checks for news every 15 minutes (96 tasks/day)
• ✅ Efficient: An agent that runs once per day and processes all findings in one task

Be Specific in Instructions

Vague prompts cause agents to explore unnecessary paths, burning credits. Be precise:

❌ Vague	✅ Specific
"Research competitors and send me findings"	"Check Crunchbase and Google News for funding announcements about [competitor list]. Send a one-paragraph summary of any new rounds."
"Monitor our social media"	"Check the Company page on LinkedIn for new posts once daily. If there's a new post, draft a Slack message quoting the first 200 characters."

Use Webhooks Instead of Polling

When possible, use inbound webhooks to trigger agents reactively rather than having agents poll on a schedule:

• Polling: Agent checks an API every 15 minutes = 96 tasks/day whether data is new or not
• Webhooks: External service pushes data → agent processes it = 0 cost until something happens

Set Clear Stopping Conditions

Tell your agent when to stop to avoid unnecessary follow-up tasks:

• "Send your report and stop — don't ask for feedback or suggest follow-ups."
• "If no results are found, log it and stop."
• "Do not propose additional research unless specifically asked."

Strategy 4: Monitor and Audit Usage

The Usage dashboard shows credit consumption per agent. Check it regularly to spot problems:

Signs of Inefficient Usage

• High per-task cost variance — Same agent sometimes costs 0.15 credits, other times 1.20+ → Check if the intelligence tier is stable
• Steady burn rate with no output — Agent using credits but not delivering → Review instructions for loops or vague prompts
• Unexpected spikes — Sudden jump in usage → Check for new schedules, peer-linked agents triggering each other, or changed instructions

Burn Rate Alerts

Gobii monitors burn rate across your agents. If consumption exceeds normal thresholds, you'll see warnings. Common causes:

• A newly created agent with no budget cap
• An agent stuck in a loop (e.g., Agent A passes work to Agent B, which passes work back to Agent A)
• A schedule set too frequently for the task

Summary: Optimization Cheat Sheet

Action	Impact	Effort
Lower intelligence tier (e.g., Max → Premium)	Up to 33% savings per task	Low — one dropdown change
Reduce schedule frequency (e.g., every 15min → hourly)	Up to 75% fewer tasks/day	Low — one setting change
Set a daily budget cap	Prevents runaway costs	Low — one slider change
Tighten agent instructions	Fewer wasted exploratory tasks	Medium — prompt refinement
Replace polling with webhooks	Zero cost when idle	Medium — integration setup
Split broad agents into focused specialists	Fewer context-switching tasks	High — agent redesign

Bottom line: Start simple, monitor usage, and iterate. Most users can cut their credit consumption by 50% or more by combining a lower intelligence tier with tighter schedules and clearer instructions.