Vascend · Talos CLI

Talos, the coding agent that sees long missions through

It reads your project, edits files and runs commands from your terminal. Pick the model, set the permissions and govern context with Epoch, Lease and Lens.

It lights up in your terminal.

One mission at a time. See it through with Talos.

  • 1
    Long missionsReads, edits and runs the tests, showing you every step.
  • 2
    The model you wantClaude, GPT, Gemini, Grok, DeepSeek or a local model.
  • 3
    Context that holdsEpoch, Lease and Lens keep the mission on track.
  • 4
    You decideThree permission levels, every change as a diff to review.
  • 5
    Runs on your computerYour keys stay yours, local models without the cloud.

The models you already use

  • Claude
  • OpenAI
  • Codex
  • Grok
  • Gemini
  • DeepSeek
  • OpenRouter
  • Ollama

An agent built around your project

Pick the model for every mission

Claude, GPT, Grok, Gemini, DeepSeek or a local model: switch with /model without changing your flow.

Context that holds

The epoch keeps the goal, leases put an expiry on focus, lenses pick what matters. /context explain shows it all.

Parallel collaborators

Split the work across agents and follow them from /agents.

Skills, MCP and extensions

AGENTS.md, skills and MCP servers for the way you work.

In your scripts too

talos -p in CI, JSONL RPC and a TypeScript SDK.

From the first command to a finished mission, Talos works in your terminal.

  1. 01

    Sign in within a minute

    Run talos, confirm the code on the Vascend site and connect the provider you already use: a Claude or ChatGPT subscription, or an API key.

  2. 02

    Describe the mission

    A concrete outcome, the constraints and how to verify it. Talos opens an epoch and drafts the task list.

  3. 03

    Talos works

    It reads files, searches the code, edits and runs the tests, showing every step as it happens.

  4. 04

    You decide

    Actions ask for confirmation according to the permissions you choose. Changes arrive as diffs to review with Ctrl+R.

Your models, on your computer. Talos in your pocket, anywhere.

Run /remote and follow Talos from the Vascend Satellis app on your phone or tablet. Local models like Ollama and LM Studio keep running on your computer: from your phone you write, approve and watch the work move forward.

  • Local models stay on your computer
  • Approve changes with a tap
  • End-to-end encrypted connection
  • The same Vascend account on every device

Connect once, work with your tools

  • Use your Claude subscription
  • Sign in with ChatGPT and Codex
  • Ask Grok
  • Local models with Ollama and LM Studio
  • MCP servers and their tools
  • Project rules in AGENTS.md
  • TypeScript SDK and RPC
  • The same engine inside Satellis

You stay in control

  • Runs on your computer, in your terminal
  • Three permission levels, switch with Shift+Tab
  • Provider keys stay yours
  • Local models, no cloud
  • Inspectable context with /context explain
  • Sign in with your Vascend account

Start with one command.

Windows, macOS and Linux. The official installers add Talos to your PATH and install Node.js when it's missing.

curl -fsSL https://vascend.it/install | bash
Installation guide