CopperTide/Architecture

How the engine works

The engine records each step of each run, saves the parts of the world a step changed, and restores any step on demand. Connectors run the systems. Checks decide the verdict.

Components

Five parts

01

Connectors

Each kind of system plugs in through its own connector. A connector starts a fresh copy of the system, carries out the agent’s actions, saves and restores its state, and compares two saved states. The engine holds no system-specific code.

02

World clock

Each world keeps simulated time. Scenario events fire on the clock, and agents receive the world’s date and time.

03

Saves and restores

After each step with a state change, the engine saves the parts the step touched. Each connector declares a restore mode: exact save, seeded replay, approximate or none.

04

Agents and runtimes

CLI mode runs Claude Code or Codex headless with only the world’s tools over MCP. API mode runs the built-in loop and records each model call. HTTP agents receive the task, an MCP endpoint and a token.

05

Checks

Goal checks and safety checks run as code against the end state. An optional AI review summarizes a run and never scores one.

Experiments

Experiments, not anecdotes

An experiment runs one to four setups on one world and scenario, N runs each, interleaved. Each setup reports pass rate with a 95% Wilson interval, checks passed, safety violations, denied actions, flags for a human, cost and time.

Nine passes in ten gives an interval from 60 to 98 percent. Thirty runs narrow the interval.

0setups per experiment, at most
0Wilson interval on each pass rate
0per run on our pharma example
0model spend for the same run

Replay and fork

Scrub to any step, branch from any step

The run view puts each agent’s steps on one timeline. Scrub to a step to see each system’s state at the time.

Fork from any step to change the model, the prompt or the team, and continue from there.

Shakedown run view of a pharma example run: fill line, QC lab, warehouse, carriers, customers and approvals at one step
A Shakedown run on our pharma example. Synthetic company and data.

State, traced

One bad write, followed downstream

The check flags the failure. The trace shows the state change in each system.

  1. 01
    Agent · setup A, run 409:12

    Starts the rush order early, using castings from batch L-2209.

  2. 02
    ERP09:12

    140 castings from L-2209 move into production.

  3. 03
    Plant10:05

    Press 2 turns the castings into 140 finished parts. Quality had flagged those castings as suspect.

  4. 04
    Quality14:40

    The batch fails inspection. All 140 parts go to quarantine.

  5. 05
    CRM15:10

    The ship date moves six days. The customer expected Friday.

  6. ✕
    Checkresult

    Failed: the agent used flagged material. Setup B never did this in ten runs.

Each figure in a report carries a source label: observed in a system at the end of the run, or derived from observed values. Simulated systems carry their own label.

SampleThe steps and numbers here form an illustration.
Next

Shakedown

→