agentprism-repl-orchestration

REPL orchestration: quickstart

Context: JavaScript sent to the MCP repl tool. This is not workflow-script code: REPL agent() takes (modelSpec, task, options?), top-level return is a syntax error, and named bindings persist between calls.

The REPL is one persistent QuickJS-in-WASM workspace per absolute projectDir. Use it when you want to inspect intermediate results and decide the next step interactively. Use workflow when the full orchestration is known up front and should be repeatable as one deterministic script.

First eval

const audit = agent("codex", "Inspect the parser for correctness bugs");

Send that code using:

{ "action": "eval", "projectDir": "/absolute/project", "code": "..." }

agent() returns a persistent promise-handle immediately. Storing the handle before awaiting preserves its id, queue(), steer(), and cancel() methods for later evals.

Inspect or await it in another eval:

agents()

const report = await audit;
report

A completed eval returns { output, result? }. If the soft hold bound expires while the eval remains suspended, the tool returns { output, running: [callIds] }; execution continues server-side. Poll without running new code by evaluating the empty string:

{ "action": "eval", "projectDir": "/absolute/project", "code": "" }

Essential semantics

• Top-level await works; top-level return does not.
• let, const, var, functions, and classes remain available to later evals.
• _ is the previous eval's completion value.
• Console output is returned as text but is not a persistent value. Assign values you need later.
• There is no filesystem, network, import, or general timer API in the VM. Subagents perform external work; sleep(ms) is the one host-backed timer.
• The default eval hold is 60 seconds and the per-call maximum is 120 seconds. This is a response hold, not cancellation.
• Use interrupt with a call id to cancel one subagent/queued turn, or omit id to break the currently running eval.
• workspace() shows bindings, in-flight calls, checkpoints, and diagnostics. agents() shows live agent lanes and queued turns.
• reset() tears the workspace down after the current eval completes.