Emily by Robson

Your models can change.
Build on what Emily learns.

Keep the decisions, preferences and project knowledge that make your work yours. Emily holds memory separately from the model. Bring in contributions from compatible connected assistants, so your next conversation has more to build on.

Start your free trial See how teaching works

Free trial $0. Companion CAD45/month before applicable taxes.

Less re-explaining.
More to build on.

A useful AI relationship needs more than a long conversation. It needs context that can outlast a session, a tool or a model.

Pick up the project

Keep the reasons behind a decision alongside the decision itself, so the next conversation has a place to start.

Bring your learning with you

Compatible connected assistants can contribute attributed information to Emily’s stored context. A contribution can remain unverified until it is reviewed.

Make the next step useful

Turn relevant context into a draft, a checklist or a better question. Check what was recalled and what is still missing before acting.

Current Persistent, attributed contributions are already possible. Saving information does not guarantee it will be recalled correctly; recall and receipt reliability improvements are in development.

Illustrative preview

Teach once.
Build on it next time.

Teaching means adding useful context to memory. It does not fine-tune the underlying model’s weights or turn a suggestion into a verified fact.

This fictional walkthrough shows the intended experience, not a recording of a completed product test. The receipt display and reliable fresh-session recall are still in development.

  1. Teach through a connected assistant

    While planning the repair café in a compatible assistant, contribute the useful decisions to Emily.

    “Keep the doors open from 10, leave a step-free route, and check with me before buying supplies.”

    Contribution: workshop brief v2. Contributor: a connected assistant.

  2. Keep an attributed receipt

    The proposed receipt makes the saved information and its review state visible.

    Purpose & relevance
    Saturday setup; access and spending decisions.
    Source & contributor
    Workshop brief v2; connected assistant.
    Time & review state
    Friday, 16:00 (example); unverified contribution.
    Uncertainty
    Volunteer arrival time is unconfirmed.

  3. Recall in a fresh session

    Open a new Emily conversation without pasting the brief again.

    “What should I do before the repair café opens?”

    The intended result retrieves the saved decisions and names their source.

  4. Get a useful next step

    An example answer turns those decisions into a reviewable plan:

    • Check the step-free route before 10.
    • Confirm when volunteers arrive.
    • Draft a supply list for approval.

    Based on workshop brief v2, still unverified. The result is a draft for you to review.

Contributions today. More control in development.

Connected MCP contributions have already produced durable, attributed, unverified records. Owner-approved teaching grants for a specific assistant, topic and Emily target are planned; expiry and revocation controls are design work, not a released feature.

Connect in a way
that fits your work.

A model that generates a reply and an assistant that contributes through a connection play different roles. Compatibility depends on the interface, permissions and deployment.

Available

Emily on the web

Start with the hosted experience. Sign in with Google, choose a plan and talk with your Emily.

Explore the trial
Available connection

Compatible MCP assistants

Use a remote MCP client that supports Emily’s browser sign-in. Connect to your own Emily; access depends on your account.

Read the MCP setup guide
Beta · access by request

EmilyRemote SDK

Build a custom interface with the EmilyRemote SDK. Current beta access and Linux setup guidance are available through Robson.

Explore developer access
Documented connection

Email

Use the Google-account email address registered with Emily. Follow the guide and confirm access for your account.

Read the email setup guide

Model choice is deployment-specific

Emily’s published integrations include OpenAI, Anthropic and Google Gemini. That does not mean every model or assistant is supported. Hosted self-service model selection and bring-your-own-key are coming soon. Ask Robson to confirm a particular model, version and workflow before relying on it.

Controlled autonomy · planned experience

Give the work direction.
Keep authority with you.

Lasting context can make delegated work more useful: the goal, the reasons behind it and the boundaries travel together.

Multi-stage autonomous work and scoped teaching permissions are in development. The planned experience lets you define the task, review its progress and decide which actions need approval.

Illustrative task boundary

“Prepare Saturday’s setup plan.”

Use
The workshop brief and confirmed decisions.
Produce
A draft checklist with sources and open questions.
Ask first
Before spending, contacting anyone or changing the plan.

Knowing something is not permission to act on it.

Know where your context
and your models run.

Where memory is stored, where a model processes it and who operates the tools are separate choices.

Available

Hosted in Canada

Emily’s hosted service stores data in Vancouver. When a third-party model is used, that provider processes the context sent to it. Canadian storage does not mean every processing step stays in Canada.

Preview

Hosted Local Mode

A reduced-capability local-model preview on Robson’s infrastructure. “Local” here does not mean Emily is installed on your own device or operates without a network.

Discuss & validate

Your infrastructure or air-gap

Organisation-controlled and isolated deployments require a deployment review. Model runtime, storage, sign-in, installation, updates and offline operation must be verified for the configuration; air-gap is not a self-service Companion feature.

Discuss a deployment with Robson

Start with a conversation.
Choose what comes next.

Try Emily, then choose Companion when you’re ready. Review your allowance, access and current terms before subscribing.

Free trial

Get to know Emily

$0

Review the current trial duration and allowance when you start.

No paid subscription starts just by choosing the trial.

Start your free trial
Available

Companion

CAD45/month

Before applicable taxes.

For ongoing conversations with your Emily. Review the current plan allowance and access in Billing.

Review Companion in Billing
Coming soon

Companion Plus

Further capabilities and capacity are planned.

Developer

A future subscription tier for building with Emily.

These plans are not available for purchase. SDK beta access is a separate request.

Free trial $0. Companion CAD45/month before applicable taxes.

Connection availability and subscription entitlements are separate. Paid add-ons and token top-ups are not currently available for purchase.

A few useful distinctions.

Does teaching Emily train the underlying model?

No. Teaching here means contributing contextual information to Emily’s memory. It does not update a provider’s model weights. Contributions can remain unverified, and a stored record is not proof of correct recall or expertise.

Can another assistant teach my Emily today?

Compatible connected assistants can contribute attributed information through MCP. That already exists. Reliable receipts and fresh-session recall are being improved; owner-approved grants limited by assistant, topic and target remain planned.

What stays when I change models?

Emily’s stored memory is separate from the underlying language model. A supported model change can use that same memory, but retrieval, answer quality and tool support can differ. Hosted self-service model selection and bring-your-own-key are coming soon.

Does the free trial turn into a paid subscription?

The trial is $0. A paid Companion subscription requires choosing the plan and completing its payment flow. Review the trial duration, allowance and current terms when you start. Existing subscribers can use Billing to manage their plan.

Can Emily work autonomously on anything I ask?

No. What Emily can do depends on connected tools, account access and the permissions for the task. Controlled multi-stage work is a direction under development, not a promise of unrestricted or guaranteed completion. Review sources, uncertainty and consequential actions.

Give your next conversation
something to build on.

Start getting to know Emily, one useful conversation at a time.

Start your free trialSign in

Free trial $0. Companion CAD45/month before applicable taxes.

What’s new