Skip to main content
$1 free credit on signup

Build & deploy your own AI agents

Spin up a dedicated agent workspace in minutes. Give it MCP tools, workflows, and a knowledge base — then let it plan, act, and get real work done. Powered by an OpenAI-compatible multimodal inference API on open models.

agent://sessionlive
$agent run "Summarize this week's support tickets"›planning · breaking the goal into steps›knowledge.search("support tickets")›mcp.github.list_issues(state="open")›workflow: weekly-digest ▸ 4 steps✓done · draft ready in /out/digest.md$
Agent Service

Your own AI agent workspace

Agent Service gives you a personal, always-on workspace with persistent storage and dedicated resources. Create as many agents inside it as you like, hand them a task, and let them plan, use tools, and run it end-to-end — all from your browser.

A dedicated workspace

Persistent storage, dedicated resources, and as many agents inside it as you like.

Your files, agents, and history stay put between sessions — nothing resets when you close the tab.

Terminal & files, built in

Enter the agent, open a real terminal, and browse its file system — right in the browser.

Agents run commands, read and write files, and use tools to actually finish the job, not just talk about it.

Bring any capable model

Back your agent with any tool-calling chat model in the catalog.

Pick a default model at setup and change it anytime — vision-capable models included.

Simple monthly plans

Solo, Pair, or Team — billed monthly from your wallet, model usage per token.

Upgrade for the pro-rated difference, downgrade for a refund, and stop anytime while keeping your files.

Enter your agent from anywhere

Subscribe to a plan, pick a model, and your workspace is live in seconds — shell, files, and skills already wired in. Connect from any device, hand it a task, and let it run. No local setup, no orchestration to babysit.

What your agent can do

Tools, workflows & knowledge — built in

Every workspace ships with the capabilities agents need to do real work — no glue code, no separate services to stitch together.

MCP Tools

Give your agent real tools through the Model Context Protocol.

Connect your own MCP servers, or use the managed gateway tools — access is governed per account so agents only see what they're allowed to call.

Learn more

Workflows

Automate multi-step jobs your agent runs on demand.

Define a workflow, run it with inputs, and track every run and step — with per-run cost visibility and the ability to cancel mid-flight.

Learn more

Knowledge Base

Ground your agent in your own documents.

Create knowledge corpora and add documents by text or URL. They're chunked automatically so your agent retrieves the right context when it answers.

Learn more

Skills

Extend agents with installable, ready-made abilities.

Native skills like image generation, editing, speech-to-text, and text-to-speech drop straight in — no plumbing required.

Learn more
Use cases

What you can build with agents

From research and coding to support and automation — agents that take the task off your plate

Research & analysis agents

Point an agent at a question and let it dig for the answer.

It browses sources, extracts and cross-checks facts, and hands back a synthesized brief with citations.

Start building

Coding & DevOps agents

Agents that read a repo, write code, and run the tools to ship it.

Shell and file access let them build, test, and iterate until the change actually works.

Start building

Customer support agents

Resolve tickets end-to-end by pulling from your own knowledge base.

Grounded in your data with retrieval, they answer, take action, and escalate only when needed.

Start building

Workflow automation agents

Automate the multi-step busywork that spans your tools and APIs.

Agents call your services on a trigger or schedule, moving work forward without a human in the loop.

Start building
The foundation underneath

One OpenAI-compatible API, every modality

Your agents — and your own code — reach every model through one endpoint, one eco_ API key, and one wallet. Try any of it live in the Playground, then drop the same request into production with the official OpenAI SDK.

Explore the API
ChatText & tool-calling
VisionImage understanding
ImageGenerate & edit
SpeechTTS, STT & voice clone
VideoAsync generation
EmbeddingsVectorize anything
RerankerScore & rerank
7model categories, one key
The platform underneath

The models behind your agents

Open models your agents call for reasoning, vision, speech, and retrieval — one OpenAI-compatible API, no vendor lock-in

Pricing

Simple, transparent pricing

Start with $1 free credit. Agent workspaces are billed monthly from your wallet — model usage is metered separately, per token.

View full pricing

Solo

A dedicated workspace for one agent — everything you need to get real work done.

$10/month
  • 1 agent workspace
  • Persistent storage & resources
  • MCP tools, workflows & knowledge
Popular

Pair

Room for two agents in one workspace — the sweet spot for most builders.

$20/month
  • 2 agents in one workspace
  • Terminal & file access
  • Change or upgrade plan anytime

Team

More capacity and agents for teams collaborating in a shared space.

$50/month
  • 5 agents in one workspace
  • Larger CPU, memory & storage
  • Shared, unified billing