Hosted AI agent workspace

Hosted AI agents that do the heavy lifting.

Windlass gives you an isolated workspace where an AI agent writes and runs code, works with your files, browses the web and uses your tools. Open it in a browser. Nothing to install, nothing to babysit.

Start free trial See pricing

7-day free trial · No card required · Cancel anytime

Project 1 · Quarterly reportWeekly usage 34%

Projects

Project 1

Project 2

Sessions

Quarterly report

Fix checkout tests

Competitor research

Import CSV into Postgres

Turn the sales export in Attachments into a quarterly report with charts. Flag anything unusual.

read Attachments/sales-q3.csv · 48,211 rows

bash: python analyze.py → summary.json

wrote report/q3-report.md · 4 charts

checking refund spike in week 37…

Report ready at Project 1/report/q3-report.md. Revenue is up 12% QoQ. One anomaly: refunds tripled in week 37 — 80% come from a single SKU. Want me to draft a note to the supplier?

Message or run a task… Send

What you get

A real environment, not a chat box

The agent has a Linux shell, a filesystem, network access and your connectors. It plans, executes, checks its own work and reports back — while you review.

Runs code and commands

Python, Node, shell, git, package managers — the agent works in its own container with full autonomy and no approval prompts.

Your files, on the server

Drop files into the chat or the Files page. Screenshots get automatic OCR so the agent can read them. Results land in your project folders.

Web and connectors

The agent fetches pages, searches and calls tools you connect through MCP: GitHub, databases, internal APIs, knowledge bases.

Long sessions that stay sharp

Up to 1M tokens of context with automatic compaction. Fork a session to explore alternatives, search your history, pick up where you left off.

Voice input

Describe a task out loud. The recording is transcribed into the message box so you can edit it before sending.

Usage you can predict

Each plan includes a weekly allowance that resets every 7 days. No five-hour windows, no daily caps, no surprise invoices.

How it works

From sign-up to delegated work in minutes

Pick a plan

Start a 7-day trial without a card. Your isolated workspace is provisioned on the spot.

Open the workspace

Create a project, attach files, connect the tools you use. Everything lives in the browser.

Delegate and review

Give the agent a task. It plans, runs, checks and reports; you read the result and steer.

Made for

Work that used to take an afternoon

Ship software

Fix failing tests, migrate a schema, write a CLI, wire an integration — with the code checked out in the workspace and the commands actually run.

Turn documents into deliverables

Spreadsheets, PDFs, screenshots and exports go in; reports, slides, cleaned data sets and summaries come out.

Research and monitoring

Competitor scans, pricing checks, literature reviews — with sources fetched live and results saved to your project.

Operate your tools

Connect a CRM, a database or a ticketing system through MCP and let the agent do the repetitive work inside them.

Pricing

Simple monthly plans

Every plan includes the full workspace. Plans differ in weekly allowance, models and storage. Compare details on the pricing page.

Starter

For individuals getting real work done with an agent.

$39/ month

Weekly usage allowance (1×)
DeepSeek V4 Flash
2 GB workspace storage
Isolated container, web access, MCP connectors
File attachments with OCR

Start free trial

Common questions

What does the agent actually run on?

Each account gets its own Linux container with a shell, a filesystem and network access. The agent's commands run there — never on your computer. Model inference is provided by DeepSeek V4 through the DeepSeek API.

How do weekly allowances work?

Your plan includes an allowance that resets every 7 days from the day your subscription started. Everything the agent processes — input, cached input, output and reasoning — counts toward it. The Overview page shows what is left and when it resets. There are no daily or five-hour caps.

What happens when I use up the allowance?

The agent pauses until the reset; your files, sessions and workspace stay available. You can upgrade at any time for a larger allowance.

Is my data used to train models?

No. Prompts and attachments are sent to the model provider to generate responses and nothing else. See the Privacy Policy for what is stored and for how long.

Can I bring my own tools?

Yes. The workspace supports MCP servers over HTTP or as local commands. Add them in the workspace settings and the agent picks them up automatically.

Can I cancel?

Anytime, from the Billing page. You keep access until the end of the paid period. The first payment of a new subscription is refundable within 14 days.