Run cost estimation

GET /runs/estimate returns a cost range for a new run before you execute it, based on the last 20 successful runs for the same team.

Why this matters

AI pipeline costs vary by request complexity, model, and team configuration. Estimation gives teams and their clients an expectation before committing:

• Developers avoid surprises on budget-constrained workspaces
• Agencies can quote clients accurately before running
• Automated systems can gate on cost before dispatching

API

GET /runs/estimate?team=DevTeam

Param	Default	Description
team	(required)	Team to estimate for
limit	20	Sample size (5–100 most recent successful runs)

{
  "team": "DevTeam",
  "based_on_runs": 18,
  "min_usd": 0.0021,
  "median_usd": 0.0038,
  "p75_usd": 0.0055,
  "max_usd": 0.0092
}

When there is no history, all cost fields return null:

{
  "team": "DevTeam",
  "based_on_runs": 0,
  "min_usd": null,
  "median_usd": null,
  "p75_usd": null,
  "max_usd": null
}

Interpretation

Field	Meaning
min_usd	Cheapest run in the sample — simple requests
median_usd	Typical cost — half of runs cost less than this
p75_usd	75th percentile — a comfortable upper bound for budgeting
max_usd	Most expensive run in the sample — complex requests

For client quoting, p75_usd gives a conservative estimate that covers most real-world runs. The median_usd is appropriate for internal cost forecasting.

Example usage

Pre-run cost gate (Python)

import httpx

client = httpx.Client(base_url="https://antcrew.org", headers={"X-Api-Key": "acw_..."})

est = client.get("/runs/estimate?team=DevTeam").json()
budget = 0.01  # $0.01 per-run budget

if est["based_on_runs"] > 0 and est["p75_usd"] and est["p75_usd"] > budget:
    raise ValueError(
        f"Estimated run cost (p75 ${est['p75_usd']:.4f}) exceeds budget ${budget:.4f}. "
        "Increase budget or use a smaller model."
    )

# Safe to proceed
client.post("/run/", json={"team": "DevTeam", "request": "..."})

Client quote tool (CLI)

cost=$(curl -s -H "X-Api-Key: acw_..." \
  "https://antcrew.org/runs/estimate?team=DevTeam" \
  | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['p75_usd'] or 'N/A')")

echo "Estimated cost for this run: \$$cost"

How the estimate is computed

The endpoint queries the last limit successful runs for the team in your workspace, sorts them by cost, and returns the distribution. Only runs with status = "success" and cost_usd > 0 are included in the sample.

The estimate improves in accuracy as your workspace accumulates more runs. based_on_runs tells you how many data points were used — treat estimates based on fewer than 5 runs as rough guidance only.