Start With Clarity

Your first task

1. Sign in with an available provider or email link. New personal trials require an eligible phone number; existing members can open their workspace.
2. Create a task for one decision. Add the customer, the problem, your current offer and the outcome you want.
3. For a structured assessment, open Roadmap → Diagnostics → Run a diagnostic. For a model-backed conversation, configure your model in LLM settings.
4. Review the result, choose one customer check and keep its outcome with the task.

First Prompt

We help small retailers track stock. Owners say stock errors matter, but few agree to a paid trial. Help me identify the assumptions to test and draft five customer interview questions. Distinguish what I have told you from what we still need to learn.

• This is an example for BYOK Chat.
• Replace it with your own context; do not invent customer evidence to complete an assessment.

Find The Gap

• Compare your reported situation or offer with customer requirements.
• See where your offer falls short and what to investigate with customers.
• Use the result to plan your next interview or test before committing more work.

1. Select one active task and open Roadmap → Diagnostics.
2. Select Run a diagnostic and enter your reported offer and customer-need scores. The form explains the 0–10 scale.
3. Add requirement detail and customer action conditions when you have the information. Unknown is not zero.
4. Review your inputs, then select Review and save. The assessment stays in the task; you can return to it or reassess later.

• A context review uses your description; a structured assessment uses scores you provide.
• These methods stay separate so their results are not mistaken for the same measurement.
• Reported scores are not verified evidence, success probabilities or execution permission.

BYOK Chat

• BYOK means Bring Your Own Key.
• Configure a supported provider and model. Compare options, use selected documents and draft plans.
• The workspace keeps the conversation with your project.

• Use a one-request key, or deliberately save an encrypted key where that option is enabled. Manage and revoke saved credentials in LLM settings.
• Add only the files and sources you need. Review generated answers and check important claims against their sources.
• Your model provider bills its usage separately. ALLYK credits cover completed billable interactions; see Pricing for the current schedule.
• Diagnostics itself needs no model key. Local model and device capabilities depend on a supported, configured Install.

• Choose a compatible API such as DeepSeek or an open-weight model provider.
• You can also host a compatible model yourself.
• In LLM settings, select OpenAI-compatible. Enter the model ID, public HTTPS API URL and key.

• The hosted app cannot connect directly to localhost or private-network model servers.
• Self-hosting still has infrastructure and maintenance costs.
• A separate consumer chat subscription is not required; your total cost is the ALLYK plan plus API usage or hosting.
• Model capability and compatibility vary by provider.

• Use the dedicated key controls.
• Never put API keys into a normal chat message.

Your Roadmap

• Overview shows evidence, decision coverage, next actions and outcomes.
• Diagnostics is a separate view for assessments and saved reviews.
• Both belong to the same Roadmap.

• The decision coverage view shows what your evidence supports and what is still uncertain.
• After a customer check, record what happened and revisit the next step.
• Switching between Overview and Diagnostics keeps your current diagnostic draft.

Plugins And API

• API and MCP: list eligible tasks, preview Diagnostics, explicitly save and read reviews.
• They do not expose ALLYK Chat, general workspace history or Agent execution.

• The ChatGPT/Codex and Claude plugins add Diagnostics to your AI conversation through the hosted ALLYK service.
• Claude supports a Claude Code package and compatible remote connections.
• The hosted service is ready for controlled testing; marketplace publication and general customer installation remain pending.

• When enabled, connect your ALLYK account through the host. Review the requested read/write scope.
• ChatGPT/Codex or Claude supplies the conversation and may charge for its own usage.
• Continue inside ALLYK when you want BYOK Chat or supported Agent workflows.

Costs and access

• Start with a no-card trial.
• The signup flow confirms its current duration and eligibility.
• Existing trial end dates remain unchanged.
• Choose Individual, Builder or Team by active users and monthly credits; the plans include the same currently released core features.

• One completed chat response or billable action uses one ALLYK credit.
• Failed or denied actions, drafts, previews and retries do not.
• Your BYOK provider may still charge for its own model usage.
• Diagnostics is included in your trial or paid plan, with no extra assessment charge. Service request limits apply.

Your Data

• Keep credentials out of messages.
• Share only content you have permission to process and that the feature supports.
• Your chosen model provider and any plugin host have their own data policies.

• A Diagnostics host sees the inputs and results exchanged with it.
• Disconnecting stops future access but cannot erase copies the host already received.
• ALLYK keeps its private calculation code on its servers.

• Archiving removes a record from active analysis; it is not deletion.
• Use Privacy settings for rights requests.
• Product improvement is optional. Workspace content is not automatically research evidence or training data.