Connection guide

Bring the model access you already use.

Limitcode supports 40+ direct APIs, enterprise cloud gateways, account sign-in, existing CLI sessions, and local OpenAI-compatible runtimes. Each route has different billing, privacy, and account-risk boundaries.

API keyDirect provider billing OAuthAccount eligibility applies CLIReuse a local login LocalYour endpoint, your runtime
01 — Direct APIs & Cloud Gateways

Use a provider-issued key.

Requests go directly from Sublime Text to the provider or gateway you select. Configure any provider via Ctrl+Alt+K or by adding its settings ID to api_keys in Limitcode.sublime-settings.

API KEY · openrouter

OpenRouter

Unified API key for 500+ models across Anthropic, OpenAI, Google, Meta, DeepSeek, and more.

API KEY · anthropic

Anthropic

Direct API key for Claude Sonnet 5, Claude Opus 5.5, and Claude Fable 5.1.

API KEY · openai

OpenAI

Direct API key for GPT-6, GPT-6 Astra, GPT-6 Sol, and reasoning series.

API KEY · gemini

Google Gemini

Google AI Studio API key for Gemini 4 Argon and Gemini 3.7 Flash.

API KEY · deepseek

DeepSeek

Direct API key (api.deepseek.com) for DeepSeek V4 and reasoning models.

API KEY · xai

xAI (Grok)

Direct API key (api.x.ai/v1) for Grok 4.20 and coding models.

API KEY · mistral

Mistral AI

Direct API key (api.mistral.ai/v1) for Mistral Large, Codestral, and Devstral.

API KEY · groq

Groq

Direct API key (api.groq.com/openai/v1) for ultra-fast LPU inference.

API KEY · cerebras

Cerebras

Direct API key (api.cerebras.ai/v1) for high-speed wafer-scale inference.

API KEY · together

Together AI

Direct API key (api.together.xyz/v1) for open-weight coding and chat models.

API KEY · deepinfra

DeepInfra

Direct API key (api.deepinfra.com/v1/openai) for serverless open models.

API KEY · moonshot

Moonshot AI / Kimi

Direct API key (api.moonshot.ai/v1) for Kimi K2.5 and long-context models.

API KEY · alibaba

Alibaba DashScope (Qwen)

DashScope API key for Qwen3 Coder and Qwen models.

API KEY · xiaomi

Xiaomi MiMo

Direct API key for Xiaomi MiMo reasoning and coding models.

API KEY · nvidia

NVIDIA NIM

Direct API key (integrate.api.nvidia.com/v1) for NVIDIA-hosted inference.

API KEY · perplexity

Perplexity

Direct API key (api.perplexity.ai) for Sonar online and reasoning models.

API KEY · venice

Venice AI

Direct API key (api.venice.ai/api/v1) for privacy-focused model inference.

API KEY · cohere

Cohere

Direct API key for Command R+ and enterprise Cohere models.

API KEY · opencode-zen / opencode-go

OpenCode Zen & Go

Curated coding gateways (opencode.ai/zen/v1 and opencode.ai/zen/go/v1).

API KEY · bedrock

Amazon Bedrock

AWS Bedrock Converse API for enterprise-hosted Claude, Llama, and Titan models.

API KEY · azure

Azure OpenAI

Microsoft Azure Cognitive Services endpoint and key.

API KEY · vertex / vertex_anthropic

Google Cloud Vertex AI

Enterprise Google Cloud Vertex endpoints for Gemini and Claude models.

API KEY · cloudflare

Cloudflare Workers AI

Serverless edge inference via Cloudflare Workers AI.

API KEY · vercel

Vercel AI Gateway

Unified routing and caching via ai-gateway.vercel.sh/v1.

API KEY · helicone

Helicone

Observability and proxy gateway via oai.hconeai.com/v1.

API KEY · gitlab / sap

GitLab AI & SAP AI Core

Enterprise GitLab AI (gitlab.com/api/v4/ai) and SAP AI Core (api.ai.sap.com/v1) endpoints.

API KEY · openai-compatible / custom

Custom OpenAI-Compatible

Any self-hosted vLLM, TGI, corporate proxy, or custom /v1/chat/completions URL.

02 — Account & OAuth

Sign in when a provider supports account access.

These integrations depend on provider-controlled accounts and service behavior. Availability, quotas, and compatibility can change independently of Limitcode.

03 — CLI sessions

Reuse an official local CLI login.

CLI bridges require the provider's command-line tool to be installed and authenticated before Limitcode starts the task.

Claude Code CLI (claude_code_cli)

Reuses an existing local claude login session from your terminal (Ctrl+Alt+E).

Codex CLI (codex_cli)

Reuses an existing local codex login session from your terminal.

CLI-backed providers may run their own internal agent loop and tools, outside some Limitcode permission controls.

04 — Local & custom

Point at your own runtime.

Local inference can remain on your machine when the endpoint and model are local. Tool calling depends on the endpoint and model implementation.

Ollama (ollama)

Default local OpenAI-compatible endpoint at http://localhost:11434/v1.

LM Studio (lm-studio)

Local server and models managed by LM Studio at http://localhost:1234/v1.

LiteLLM (litellm)

Local or remote proxy gateway at http://localhost:4000/v1.

Custom endpoint (openai-compatible)

User-supplied OpenAI-compatible base URL in provider_base_urls.

Before connecting

Know where the request goes.

Prompts and any file context or tool results included in a request are sent to the selected provider or endpoint. Credentials are stored in Sublime Text settings, not an operating-system keychain. Review the provider's terms before connecting a work or primary personal account.

Return to Limitcode Pro →