Skip to content

Carbon & Price Aware API · Early access

Run AI when the grid runs green.

One API that tells your agents when and where electricity is cleanest, and what a run will cost before it starts. Move the work, keep the deadline, and cut its carbon by up to 50%.

Grid forecast · next 24 hgCO₂e per kWh
Deadline 08:00Best start
Now · 16:00
238 g
Best · 02:00
123 g
Saving
−48%
Illustrative forecast for a 45-minute agent run due by 08:00.
  • Grid data every hourLive and forecast carbon intensity across Europe and North America.

  • Runs wait for clean powerGive a deadline, not a start time. The API picks the window.

  • A price on every runToken prices per model, before the first token is spent.

01 / 04 · The API

An API for when, where and how much.

The Carbon & Price Aware API is a REST and MCP service that answers three questions before any job runs: how clean the electricity is, when and where it will be cleanest before your deadline, and what the run will cost in tokens. Your code or your agent decides; the API supplies the facts.

  • GET /carbon

    How clean is it now?

    Carbon intensity for any location, measured and forecast hour by hour.

  • GET /best-window

    When should it run?

    Send a duration and a deadline. Get the cleanest start time and the saving against running now.

  • GET /best-location

    Where should it run?

    Compare regions and send flexible work to the grid with the cleanest power.

  • GET /prices

    What will it cost?

    Up-to-date token prices per model, so every run carries a price tag.

RequestREST · or as MCP tools
curl "$CARBON_API/best-window\
  ?duration_minutes=45\
  &deadline=2026-11-02T08:00:00+01:00\
  &lat=52.37&lon=4.90&power_kw=0.06" \
  -H "Authorization: Bearer $GREENPT_KEY"


          {
          
              "run_now": false,
            
              "start": "2026-11-02T01:00:00Z",
            
              "end": "2026-11-02T01:45:00Z",
            
              "gco2_kwh": 123,
            
              "if_run_now": 238,
            
              "saving_pct": 48.3,
            
              "grams_saved": 5.2
            
          }
        
Example response. Agents can call the same endpoints as tools over MCP.

02 / 04 · How the grid works

Same plug. A different grid every hour.

The grid has to match supply and demand every second. Wind, sun, hydro and nuclear are used first because their running cost is lowest. Whatever they can't cover is filled by gas and coal plants, which ramp up and down all day. So the carbon behind one kilowatt-hour can more than double between night and evening.

Where one day's electricity comes from

  • Hydro & nuclear
  • Wind
  • Solar
  • Gas
  • Coal
  • Carbon intensity
02:00 · 123 g/kWh18:00 · 303 g/kWh
Illustrative day. Real mixes differ by country, season and weather, which is why the API forecasts every hour.
  1. Night

    Wind carries the load

    Demand is low and wind often blows hardest. Fossil plants throttle down and every kWh is at its cleanest.

  2. Midday

    The sun steps in

    Solar output peaks and pushes gas off the grid, opening a second clean window on sunny days.

  3. Evening

    Fossil fills the gap

    The sun sets as everyone gets home. Gas and coal ramp up to cover the peak, and intensity climbs.

03 / 04 · Carbon-aware agents

Agents that wait for clean power.

A carbon-aware agent gets a deadline instead of a start time. It asks the API for the cleanest window before that deadline, holds its run, and starts when the grid is greenest. Nightly research, weekly reports, batch summaries, data enrichment and evals rarely need to run this very minute.

50%Up to half the CO₂e of the same run, by moving it to the cleanest hours before its deadline.

One run, shifted

  1. 16:00

    Run requested

    45 minutes of work, due by 08:00. Grid now: 238 g/kWh.

  2. 16:00 → 02:00

    Held in the queue

    The evening peak passes. Gas and coal ramp down as the wind picks up.

  3. 02:00

    Starts in the green window

    Grid at 123 g/kWh: 48% less carbon than starting at 16:00.

  4. 02:45

    Done, hours early

    Results are waiting before anyone starts their day.

  • Never misses a deadline

    The window always ends in time. If no hour is cleaner than now, the run starts now.

  • Fails open

    Missing data never blocks work. Without a forecast, the run simply goes ahead.

  • Any agent framework

    Call it over REST from your scheduler, or give your agent the MCP tools and let it plan.

04 / 04 · Price aware

Know the price of a run before it runs.

The API keeps token prices per model up to date. Combine them with your expected tokens and every run gets a price before it starts, and an exact cost when it ends.

  1. 01

    Estimate up front

    Quote a job to a customer or a manager before a single token is spent.

  2. 02

    Cap the budget

    Give each agent a ceiling per run. Stop, or switch to a smaller model, before it is crossed.

  3. 03

    Pick the right model for the money

    Compare what the same task costs on each model and route by price.

  4. 04

    Charge back per run

    Pass exact costs on to clients, teams or departments, run by run.

  5. 05

    Report carbon and cost together

    One line per run with euros, watt-hours and grams of CO₂e, ready for your sustainability report.

Run receipt

Weekly market report

Model
green-l
Started
02:00 · held 10 h
Grid
123 gCO₂e/kWh
Tokens
182,400 in · 24,900 out
Energy
41 Wh
Emissions
5.0 g CO₂e (9.8 g at 16:00)

Price of this run€0.066

Example run. The price uses today's green-l rates; energy and grid figures are illustrative.

Early access

Be first on the green window.

We're opening the Carbon & Price Aware API to a small group of teams first. Tell us what you would run, and we'll send you a key.

  • An API key for REST and MCP
  • A direct line to the team building it
  • A say in which regions and endpoints come next

Request early access

Questions, answered

Does waiting make my agent slower?

Only work you mark as flexible waits. You set the deadline, and the run always finishes before it. Interactive chats and urgent jobs run straight away.

What if there's no cleaner hour before my deadline?

Then the API answers run_now: true and your job starts immediately. The same happens when a forecast is missing, so data gaps never hold up your work.

Which regions are covered?

Grids across Europe and North America, measured and forecast every hour. Query by coordinates or by data-centre region.

How is carbon intensity calculated?

From each hour's generation mix, using lifecycle emission factors per source by default. Direct (combustion-only) factors are available too.

How does it relate to the GreenPT API?

It works on its own with any model or workload. Combined with the GreenPT API, your runs report energy and CO₂e per request out of the box.

Explore the GreenPT API →