executespec-core

Shared TypeScript client core for ExecuteSpec CLI and VS Code clients.

This package contains the typed REST/SSE client, browser-login helpers, PAT token handling, provider helpers, chat/session primitives, and local tool-runtime contracts used by the public clients. It does not contain the ExecuteSpec agent loop, prompts, model routing, scoring, billing, or project execution logic. Those remain server-side.

Install

npm install executespec-core

Default Host

The default backend is:

https://dev.executespec.ai

Consumers can override it with EXECUTESPEC_HOST, stored client config, or an explicit host option.

Usage

import {createClient, resolveConfig} from 'executespec-core';

const config = resolveConfig();
const client = createClient({host: config.host, token: 'es_pat_...'});
const me = await client.getMe();

Runtime Model

ExecuteSpec clients are thin clients:

• local clients hold credentials and optional local tool execution state
• the backend owns the agent loop, plan gate, model routing, scoring, and run history
• managed-AI credit state is read from /api/billing/managed-ai/*; unlimited BYOK/Team/Enterprise tiers use unlimitedAiCredits=true and -1 included/remaining credit sentinels
• local-hands execution is opt-in and guarded by the caller surface
• local-hands TOOL_CALL events are executed sequentially: the client runs the local tool, posts the result, then resumes SSE reading so the backend cannot be left waiting on a stranded pending call
• app-shell continuity state includes the launcher route used by the CLI TUI as the workflow hub before entering runs, evidence, diff, permissions, settings, or session panes

License

Proprietary. All rights reserved.