Understands the page. Acts only with your approval.
Grounded in your data, aware of the page your user is viewing, and able to take safe actions through preview and confirmation. Connect WordPress, React, JavaScript, or your own backend.
No credit card · live in 10 minutes once your workspace is provisioned
Connects where your users already are
Not another chat box.
A controlled action layer.
Three things make Context Assistant different from a generic AI chat widget.
Answers, then acts — safely
Reads run automatically. Every change is shown to you first, and deletes ask for a second confirmation.
getruns automaticallyupdatepreview → Applydeletedouble confirmationKnows what's open right now
The current product, page and selected object are passed as small structured facts — no hidden DOM scraping.
type: "product",
id: 184, stock: 12
}
No data locked in one cloud
Open Remote Tools contract, one-command conversation export, and a working self-hosted core.
How it works
One brain, reached through your page, grounded in your data, gated by confirmation.
Sends allowed context — not the page body.
Hybrid RAG over your knowledge ↔ host Remote Tools.
Confirmation → action → action log.
WordPress, script, React, JS or Remote Tools.
Choose the knowledge and tools it may use.
Every executed action is recorded and reviewable.
Useful to visitors and owners alike
The same assistant helps a shopper decide — and helps the owner run the store safely.
Grounded product help
A shopper asks about the product they're viewing. The assistant uses the current item's specs and recommends only in-catalog items with real price and availability.
- Uses page context, not guesses
- Never invents products or stock
- Answers cite your own content
Safe owner actions
The owner asks for a stock update. The assistant proposes the change, waits for Apply, then writes an immutable audit entry — with the owner's real permissions.
- Preview before any write
- Runs with the owner's host permissions
- Recorded in the action log
Where teams put it to work
Spec-based recommendations, availability, order status, and safe product updates.
Answers grounded in your posts, search, and work with drafts and comments.
Help that accounts for the current screen and the selected object.
Typed actions exposed through the Remote Tools contract.
Built so the secret never leaves the server
The browser only ever holds a short-lived token. Every request is scoped to a single tenant and workspace.
- The LLM provider secret stays on the server; the embed gets a short-lived token.
- Every database query is bound to your tenant and workspace.
- The host re-checks the user's real permissions before any action.
- Delete needs double confirmation and is recoverable from trash.
Simple, honest pricing
Launch pricing — shown ahead of public self-service. Safe actions are included from Starter up.
- 100 AI replies / mo
- 1 assistant · 1 site
- 2 MB indexed text
- 7-day history
- Community support
- 1,000 AI replies / mo
- 1 assistant · 2 sites
- 10 MB indexed text
- 30-day history
- Email support
- 5,000 AI replies / mo
- 5 assistants · 10 sites
- 50 MB indexed text
- 180-day history
- Remove branding · priority email
- 20,000 AI replies / mo
- 20 assistants · 50 sites
- 200 MB indexed text
- 365-day history
- White-label · onboarding
Need full control over infrastructure? Self-hosting the open core and enterprise SSO/SLA arrangements are available — Request early access to talk to us. Compare all features →
Your data stays portable
Stream a whole workspace's history out as NDJSON, anytime.
Delete dialogs and use the WordPress privacy integration — built in, not promised.
One Remote Tools contract works with the cloud and a self-hosted core.
Frequently asked questions
How is Context Assistant different from a regular chatbot?
It knows the page a user is on and can take typed actions — but only reads run automatically. Any change is previewed and confirmed before it happens.
Does my API key end up in the browser?
No. The permanent secret stays on the server. The embed receives only a short-lived token issued by your own token endpoint.
What page data is sent?
Only an allowlist of small structured facts (like page type and product id). The full DOM or page body is never sent automatically.
Can the assistant change or delete data on its own?
No. Creates and updates require a preview and an explicit Apply; deletes need a second confirmation and are recoverable.
How do I get access today?
Public self-service signup isn't live yet. Request early access and we'll set up your workspace.
Can I use my own server?
Yes. The core is open-core and runs with Docker Compose. The same Remote Tools contract works against the cloud or your self-hosted core.
Does it work with WordPress / WooCommerce?
Yes — live-verified with WordPress and WooCommerce: 6 WordPress and 4 WooCommerce permission-aware tools today (10 total). The wordpress.org catalog listing is still in progress.
Which LLM providers are available now?
OpenAI and OpenAI-compatible providers work today. Other providers are adapters on the roadmap, not yet user-selectable options.
Can I export conversations?
Yes. A whole workspace's conversations export as streaming NDJSON, and visitor dialogs can be deleted.
Do I need a credit card for Free?
No. Free needs no card. You get 100 AI replies per month and can upgrade whenever you're ready.