Agent Cost Log
Cloud for Botsinfrastructure for your bots

Know what every run cost. Before the bill does.

Send one line per model or tool call, or your OpenTelemetry export. Get cost per run, per tool, per model and per day. No evals, no prompts, no proxy. For solo builders who only need the cost split.

Start free, no signupAdd the MCP server

500 calls/day free · Pro $9/mo · we never guess a price

run research-42 · cap $0.30
run total$0.000
cap
    plan_step$0
    summarize$0
    web_search$0
    browser.click$0
    log_costs · waiting for calls
    Sample: Logged 3 call(s) costing $0.121. Run research-42 total: $0.121. Cap crossed: would alert (watch-only; Pro posts to your webhook).
    Works with the agents you already run
    • AClaude
    • {}Codex
    • CChatGPT
    • GGrok Bot
    Remote MCP · REST · bring your own model

    How it works

    Send the cost. See the split. Catch the spike.

    01log_costs

    After each model or tool call (or once with your OTel export), send what it cost.

    02cost_report

    See spend per run, per tool, per model and per day.

    03cap alert

    Pro posts to your webhook the first time a run or a day crosses its cap.

    Honest comparison

    For solo builders who only need the cost split

    Full observability platforms do much more (traces, evals, prompts, teams), and each has a free tier. If all you need is cost per run, tool, model and day, here is what their paid plans list today.

    Langfuse Core$29/mopricing page ↗AgentOps Profrom $40/mopricing page ↗Helicone Pro$79/mo + usagepricing page ↗
    Agent Cost Log Pro$9/mocost split + cap alerts only

    List prices checked Oct 15, 2026 on each vendor's pricing page. Agent Cost Log does not replace them.

    Pricing

    Watch free. Pay when it acts.

    Free resets daily and is not a trial. Pro is $9 a month.

    Free · watch-only

    $0

    Watch-only mode. 500 calls a day: log_costs totals each run and day, cost_report shows 7 days, and runs over your cap come back as would_alert. Nothing is sent.

    • Cost per run, tool, model, day
    • 7-day report
    • would_alert when a cap is crossed
    • Sends nothing
    Acting

    Pro

    $9/month

    Acting: when a run or a day crosses its cap, one alert per day goes to your webhook. 30-day history per run, tool, model and day. 50,000 calls a day.

    • Cap alert to your webhook, once a day
    • 30-day history
    • 50,000 calls a day
    • Key instantly after checkout, cancel anytime
    Upgrade to Pro · $9/mo

    Worked example

    Your research agent makes about 40 model and tool calls per run, 25 runs a day. Free mode logs them and the weekly report shows which run and which tool cost most. With Pro at $9/month, the day a retry loop pushes a run past your $2 cap, your webhook gets the alert that day, and you can look back 30 days.

    Free (watch)
    weekly report, would_alert
    Pro (acting)
    alert the same day

    For agents

    Tools your agent gets

    Each result includes understood, job and a one-line outcome your agent can report back in plain words.

    log_costs

    Log what your agent's model and tool calls cost, then get the running total for the run and for today. Pass events (one per call, up to 500) or an OpenTelemetry OTLP/JSON trace export as otel (gen_ai.usage.input_tokens / output_tokens / gen_ai.request.model are read; trace id becomes run_id).

    cost_report

    Cost per run, per tool, per model and per day. Without an API key pass run_id (one run).

    weekly_report

    Plain-English summary of the last 7 days of spend: total, top runs, top tools, top models, busiest day. Read-only; sends nothing.

    usage_and_upgrade

    Usage today + upgrade link (not metered).

    Connect

    One line to install

    claude mcp add --transport http agent-cost-log https://costlog.cloudforbots.com/mcp
    MCP URL  https://costlog.cloudforbots.com/mcp
REST     POST https://costlog.cloudforbots.com/api/run/<tool>
Auth     optional  Authorization: Bearer <key>

    Codex, ChatGPT, Grok Bot: add a remote MCP server with the URL above. Skill file: SKILL.md · catalog entry: tool.json

    FAQ

    Questions

    What problem does this solve?

    Agent bills arrive as one number per provider. Agent Cost Log splits it by run, tool, model and day, so you can see which run or which step burned the money and cap it.

    How do I send costs?

    Call log_costs with one event per model or tool call: cost_usd if your provider returns it, or tokens plus your own per-million prices. Or pass an OpenTelemetry OTLP/JSON trace export as otel; gen_ai.usage token counts and the model are read, and the trace id becomes the run.

    Do you guess model prices?

    No. We never look up or guess a price. You send cost_usd or your own per-million prices from your provider's price page. Calls with tokens and no price are counted and flagged as unpriced.

    What is watch-only mode?

    The free plan logs and reports 7 days. When a run or a day crosses your cap it returns would_alert and sends nothing.

    What does Pro do?

    Posts one alert per day to your webhook when a run or a day crosses its cap, keeps 30 days of history per run, tool, model and day, and allows 50,000 calls a day.

    How does it compare to Langfuse, Helicone or AgentOps?

    Those are full observability platforms with traces, evals, prompts and more (Langfuse Core $29/mo, Helicone Pro $79/mo plus usage, AgentOps Pro from $40/mo, each with a free tier). Agent Cost Log is for solo builders who only need cost per run, tool and day. It does not replace them.

    Is there a model inside?

    No. Bring your own model. Agent Cost Log only adds up the numbers you send.

    What does it cost?

    Free: watch-only, 500 calls a day, 7 days of history. Pro is $9 a month. It is also included in the Cloud for Bots All Access bundle.