Cost Management

Understand what drives workflow costs and how to monitor and optimize spending.

Understanding Workflow Costs

Workflow runs consume model usage on your plan, metered like the rest of your Odeus usage. The main cost drivers are:

AI Agent Nodes

The biggest expense in most workflows. Costs depend on:

• Model used: Larger models cost more than smaller ones
• Input length: How much data you send to the agent
• Output length: How much the agent generates
• Tool usage: Web searches, Knowledge Hub retrieval, and connector actions add work

Other Nodes

• Action nodes: No model usage of their own (they call your connected apps)
• Condition / Loop / Parallel / Delay / Transform nodes: No model usage — just logic and flow control
• Approval nodes: No model usage — they pause for human review

A Loop or Parallel node that calls an Agent node many times multiplies the agent cost by the number of iterations or branches.

Monitoring Costs

Per-Node Costs

After a workflow run, each node that used a model shows its cost. Open the node to see:

• The cost for that node
• A per-model breakdown with input/output tokens

Per-Run Costs

View costs in the Runs tab:

1. Click on any run in the history
2. See the total execution cost
3. Click individual nodes to see their cost breakdown

Workflow Run History

The Runs tab also lets you review past executions — when they ran, how long they took, whether they succeeded or failed, and what each one cost — so you can spot expensive patterns over time.

Plan Usage

Workflow usage counts toward your Odeus plan's usage, billed through Stripe like the rest of the platform. Workspace admins can review overall usage and plan limits in the admin area.

Optimization Strategies

Choose the Right Model

Don't use premium models for simple tasks. Pin a smaller, faster model on agent nodes that only do lightweight extraction or classification, and reserve larger models for complex reasoning.

Task	Model Choice
Extract a field from text	Smaller/faster model ✅
Complex reasoning	Larger model ✅
Reshaping data	Transform node (no model) ✅

Use Transform Nodes for Plain Data Work

Reshaping, filtering, and formatting data don't need an LLM. Use a Transform node instead of an Agent node for:

• Filtering or sorting lists
• Picking specific fields out of a payload
• Reformatting values between steps

Filter Before You Call AI

Narrow large inputs before they reach an Agent node so you process only what matters:

Trigger (100 items) → Transform: keep the 20 relevant items
                    → Loop → Agent: process 20 items (not 100)

Optimize Agent Prompts

Shorter, clearer prompts cost less and work better:

Analyze this feedback. Return:
- Sentiment: positive/neutral/negative
- Urgency: low/medium/high
- Key issue (1 sentence)

Feedback: {{trigger.output.message}}

Use structured outputs. They're more reliable and prevent the model from generating unnecessary explanatory text.

Estimating Costs Before Launch

1. Count expected runs — form submissions, scheduled runs per day × 30, or webhook events per month.
2. Test with real data — run a handful of test runs and check the per-run cost shown in the Runs tab, then multiply by your expected volume and add a buffer.

Next Steps

• Agent Node — Understand agent node costs and optimization

• Node Types — See the full set of available nodes