Beam Credit Analysis

If URL contains enterprise.beam, the script auto-detects and uses https://api.enterprise.beamstudio.ai.

Step 2: Check for Model Overrides

Ask if any models in the agent have been changed from their defaults. Common scenario: a node originally using Gemini 3 Pro was switched to GPT 5.2.

If overrides exist, pass them to the script as --model-override OLD=NEW.

Example: --model-override GEMINI_3_PRO=GPT_5_2

Step 3: Verify Configuration

Check that BEAM_API_KEY exists in .env:

grep "BEAM_API_KEY" .env

If missing, ask user to provide it and add to .env.

Skip this step if user provides a --graph-file instead of a URL.

Step 4: Run Analysis Script

Execute the analysis script:

From URL:

python3 03-skills/beam-credit-analysis/scripts/analyze_agent_credits.py "AGENT_URL" --model-override GEMINI_3_PRO=GPT_5_2

From local graph file:

python3 03-skills/beam-credit-analysis/scripts/analyze_agent_credits.py --graph-file ./path/to/graph.json

With custom output directory:

python3 03-skills/beam-credit-analysis/scripts/analyze_agent_credits.py "AGENT_URL" --output ./custom/path

For JSON output (programmatic):

python3 03-skills/beam-credit-analysis/scripts/analyze_agent_credits.py "AGENT_URL" --json

Step 5: Review Output

The script generates a markdown file containing:

1. Credit Rates — Rates used for calculation with source reference
2. Architecture Overview — Branch descriptions and mermaid flow diagram
3. Per-Branch Tables (for each branch):
   • Node Count — Flash, Premium, Integration, Trigger counts per path
   • Credit Consumption — Credits per node type per path
   • Cost — Dollar cost per node type per path (using actual model rates from Notion pricing)
4. Summary — All paths with total nodes, credits, and cost
5. Eval Impact — Credits and cost if eval is enabled on all GPT nodes
6. Step-by-Step Breakdowns — Top 3 most expensive paths with per-node detail
7. Node Inventory — All Custom GPT and Integration nodes with models and rates

Step 6: Present Results to User

Show the user:

• File path of the generated analysis
• Summary table (all paths with total nodes, credits, cost)
• The mermaid diagram (renders in GitHub markdown)
• Highlight the cheapest and most expensive paths
• Note any assumptions (e.g., GPT 5.2 pricing estimated from GPT 5)

Script Reference

analyze_agent_credits.py

Arguments:

*One of: URL, workspace+agent IDs, or graph-file is required.

Exit codes:

• 0 = Success
• 1 = Error (auth, network, missing config)

Pricing Updates

The pricing reference is stored at references/pricing.json. To update:

1. Check the Beam Credits Notion page for changes
2. Update references/pricing.json with new rates
3. Update the last_updated field in _meta

The script supports both UPPER_CASE and lower-case model name formats for flexibility.

Error Handling

Resources

scripts/

• analyze_agent_credits.py — Main analysis engine (fetch, trace, calculate, generate)

references/

• pricing.json — Beam credit rates per model and node type (from Notion)

Related Skills

• get-beam-agent-graph — Fetch and save agent graph JSON (used internally)
• calculate-beam-agent-pricing — Design node architecture and calculate pricing from requirements