Setup · 2 minutes

Connect any editor

Taksha speaks the OpenAI API. Set a base URL + your key, choose a model — done. Works with our VS Code extension — plus Zed, Aider, SDKs, curl: any OpenAI-compatible client.

Get your key ⤓ Download PDF guide
The three things you need
Base URL
https://app.agentanywhere.ai/agents/coder/v1
API key
sk-sov-live-… (from your dashboard)
Model ids
taksha-yantra · taksha-pro · taksha-max

No key yet? Redeem an invite to get one instantly. Tip: pick Yantra (model id taksha-yantra) for agentic work, Pro for sharpest single-shot generation.

Recommended · one-click
🧩 Taksha for VS Code

Our branded extension — pre-pointed at Taksha. Install it, paste your key, and code. No base-URL or model fiddling.

⤓ Download .vsix
  1. Download the .vsix above.
  2. In VS Code: Extensions panel ▸ ⋯ (top-right) ▸ Install from VSIX… — or run code --install-extension taksha.vsix.
  3. Open Taksha in the sidebar ▸ paste your API key (get one). Base URL & model are pre-set.
  4. Pick a model — Yantra for agentic tasks (the default), Pro for single-shot generation.
ZedZed (verified today)
  1. Paste the block below into ~/.config/zed/settings.json (merge with what you have).
  2. Set your key EITHER in Zed › Settings › AI › LLM Providers › OpenAI › API Key (stored in your keychain), OR as the OPENAI_API_KEY environment variable — then fully restart Zed (Cmd+Q).
  3. Agent Panel (Cmd+?) › model picker › OpenAI section › Taksha Yantra. Agentic edits with diff review work out of the box.
{ "language_models": { "openai": { "api_url": "https://app.agentanywhere.ai/agents/coder/v1", "available_models": [ { "name": "taksha-yantra", "display_name": "Taksha Yantra (agentic)", "max_tokens": 98304, "max_output_tokens": 8000 }, { "name": "taksha-pro", "display_name": "Taksha Pro (codegen)", "max_tokens": 98304, "max_output_tokens": 8000 }, { "name": "taksha-adhunik", "display_name": "Taksha Adhunik (migration)", "max_tokens": 98304, "max_output_tokens": 8000 }, { "name": "taksha-pariksha", "display_name": "Taksha Pariksha (testing)", "max_tokens": 98304, "max_output_tokens": 8000 } ] } }, "agent": { "default_model": { "provider": "openai", "model": "taksha-yantra" } } }
CLIAider (CLI)
  1. Set the two environment variables, then launch aider with the Taksha model.
export OPENAI_API_BASE=https://app.agentanywhere.ai/agents/coder/v1 export OPENAI_API_KEY=sk-sov-live-… aider --model openai/taksha-yantra
AnyAny OpenAI-compatible app / SDK
  1. Anything that speaks the OpenAI API works — OpenAI SDKs, LangChain, Neovim (avante/codecompanion), curl.
  2. Point the base URL + key at Taksha and use model id taksha-yantra.
export OPENAI_BASE_URL=https://app.agentanywhere.ai/agents/coder/v1 export OPENAI_API_KEY=sk-sov-live-… curl $OPENAI_BASE_URL/chat/completions \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"taksha-yantra","messages":[{"role":"user","content":"hi"}]}'
Good to know
  • Streaming works out of the box (set stream: true).
  • Every response returns X-Credits-Used / X-Credits-Remaining so you can watch your balance.
  • A locked model returns a clear message telling you which plan unlocks it.
  • Your code is served from our own model, in-region — it never leaves to a third-party cloud.