CopperTide/FAQ

Questions and answers

Answers grouped by topic. Ask us anything else at hello@coppertide.ai.

Getting started

First projects

Who joins the project?

The engineer who owns the agent, plus an expert for each system involved.

What does a first project cost?

We quote after the first call. Price depends on how many systems the workflow touches.

What do we keep?

The world definition, scenarios, checks, a suite to rerun on each change, and traces of each run.

Engine and agents

How the engine works

Which agent frameworks work?

Any framework able to call MCP tools or answer an HTTP request. Claude Code and Codex run headless out of the box.

Which models?

OpenRouter, OpenAI, the Claude API, or a local OpenAI-compatible server such as Ollama, vLLM or LM Studio.

What happens on a fork when a system lacks exact saves?

The engine replays the system from its seed. If the connector supports neither, the fork refuses and names the system.

Where does run data live?

In Postgres, inside your deployment. Runs export as JSONL with a schema.

Security and data

Where data lives and moves

Does a run reach production systems?

No. Each run uses fresh instances started for the run.

Does Shakedown need internet access?

No, with a self-hosted model.

Where do model calls go?

To the provider configured per agent. A self-hosted model keeps prompts inside your network.

Back to

Architecture

→