EchoAI

EchoAI is a local-first AI coding and automation platform that combines a programmable CLI agent runtime, desktop/mobile command center, hosted model-credit cloud, and multi-channel automation.

The core product is a CLI/runtime that can run model-backed coding sessions, manage tools, permissions, memory, sessions, background tasks, MCP tools, provider routing, diagnostics, and evals.

Main Product Surfaces

• echoai CLI for chat, config, providers, agents, login, usage, MCP, sessions, security review, tasks, evals, ACP, analytics, models, memory, channels, and skills.
• Private Next.js command center for chat, runtime events, models/usage, projects, files, knowledge, notes, memories, tools, MCP, integrations, skills, automations, outputs, billing, devices, admin tickets, and settings.
• Electron desktop app with gateway pairing, runtime sessions, files, terminal tasks, browser/tooling, devices, channels, automations, memory, skills, MCP, settings, updater, and secure IPC.
• React Native and native Android/iOS/macOS control surfaces for pairing, gateway sessions, chat, approvals, voice wake/talk mode, camera/location/screen/SMS capabilities, and remote-control flows.
• Hosted cloud for admin, users, credit grants, blocking, audit logs, device authorization, usage/billing, provider readiness, and metrics.
• Astro/Starlight docs and marketing site with pricing, security, downloads, docs, account/app pages, and product positioning.

Complexity Handled

• Device-code auth, refresh tokens, HMAC/JWT-style access tokens, signed local web cookies, hosted Auth.js GitHub/Google login, internal admin headers, workspace members/roles, and audit logs.
• LemonSqueezy checkout/webhooks, subscriptions, top-ups, credit ledger, usage events, hourly credit limits, and model-cost accounting.
• Streaming chat completions, runtime event streams, WebSocket/JSON-RPC gateway, background task logs, tool-call events, approval flows, Android/macOS gateway sessions, and channel messaging.
• AgentKernel, tool registry, permission resolver, MCP tools, skills, memory, projects, knowledge files, scheduled automations, eval harness, model filters/fallbacks, and background tasks.
• User blocking, credit grants, metrics, audit logs, provider readiness, tickets surface, device pairing/revocation, billing/usage views.
• Device authorization, provider-key vaulting, permission moderation, command risk classification, and tool approvals.
• Usage ledger, metrics endpoint, analytics/export managers, eval reports, model usage, billing summaries, and security-review heuristics.
• GitHub Actions CI/release, npm package scripts, Electron packaging, hosted Hono API deployment docs, Vercel/Node hosting path, Neon/Postgres/Drizzle schema, metrics, health/readiness endpoints.

Caveats

Not every surface is equally productionized. Some mobile/native and integration surfaces are broader than they are hardened. Full root tests were noisy because checked-in samples also ran, while the focused EchoAI suite passed.

What This Demonstrates

EchoAI demonstrates senior-level AI product engineering: agent runtime design, local-first security, model/provider abstraction, SaaS billing, cross-platform desktop/mobile thinking, tool orchestration, permissions, evals, deployment, and integration-heavy architecture.

The strongest signal is the connected platform shape: CLI runtime, hosted API, billing ledger, desktop gateway, web command center, native device layer, and channel automation.

Short Version

Built EchoAI, a local-first AI coding and automation platform spanning a CLI agent runtime, desktop command center, private web dashboard, hosted model-credit API, mobile/native control surfaces, and messaging-channel integrations. The system includes tool orchestration, permission/risk controls, MCP/ACP support, model provider routing, BYOK keys, usage billing, credit ledger, streaming events, background tasks, memory, evals, diagnostics, device pairing, and admin/billing operations.