Multi-API Studio CLI

Windows-first full-screen CLI/TUI for working with multiple AI API providers.

One-command install (global)

npm install -g @svlktv/al-in-one-cli
aiocli

How it works:

• npm installs global command aiocli.
• postinstall downloads Windows binary from GitHub Releases for the package version.
• No manual PATH edits are required for users (npm handles global bin shims).

Features

• Full-screen Bubble Tea interface with premium dark styling.
• Service picker (arrow keys + Enter):
  • Nvidia Build API
  • you.com API
  • Cerebras API
  • Omniroute API
  • Agentrouter API
• 3-step reachability checks before auth:
  • DNS resolve
  • TCP connect
  • HTTP health
• API key secure storage in Windows Credential Manager.
• Dynamic model loading from selected provider.
• Streaming chat with thinking indicator.
• Reasoning text pane (shown only if provider emits reasoning deltas).
• Top metrics strip:
  • model
  • author (svlktv)
  • token usage
  • request limit + overall limit
  • current workspace
  • remaining context
• Local history persistence in SQLite by service+model sessions.

Provider mapping

• Nvidia Build API: OpenAI-compatible /v1/models + /v1/chat/completions.
• Cerebras: OpenAI-compatible base (https://api.cerebras.ai/v1) via /v1/models + /v1/chat/completions.
• AgentRouter: OpenAI-chat provider on https://agentrouter.org/v1.
• OmniRoute: OpenAI-compatible endpoints (/v1/models, /v1/chat/completions), default local URL http://localhost:20128.
• You.com: Search API (GET /v1/search, header X-API-Key) adapted into chat-like responses.

Local development run

go run ./cmd/multi-api-studio

Version info:

go run ./cmd/multi-api-studio version

Release process

1. Update package.json version.
2. Push git tag vX.Y.Z.
3. GitHub Action builds and uploads aiocli-windows-amd64.exe to the release.
4. Publish npm package (npm publish --access public) or via publish-npm workflow.

Configuration

App stores preferences in:

• %APPDATA%/multi-api-studio/config.json

Overrides for provider base URLs:

• MAS_NVIDIA_BASE_URL
• MAS_YOU_BASE_URL
• MAS_CEREBRAS_BASE_URL
• MAS_OMNIROUTE_BASE_URL
• MAS_AGENTROUTER_BASE_URL

Source docs used

• Cerebras: https://inference-docs.cerebras.ai/introduction
• AgentRouter: https://docs.agentrouter.org/en/start.html and https://docs.agentrouter.org/en/codex.html
• Nvidia: https://docs.api.nvidia.com/nim/reference/models-1
• OmniRoute: https://github.com/diegosouzapw/OmniRoute/blob/main/docs/API_REFERENCE.md
• You.com: https://you.com/docs/api-reference/search/v1-search

Notes

• Where providers do not expose exact counters, token/limit/context metrics are marked as approximate.
• Per-provider overrides (headers, routes, payloads) can be extended in dedicated adapters.