> ## Documentation Index > Fetch the complete documentation index at: https://docs.stateset.com/llms.txt > Use this file to discover all available pages before exploring further. # From Signup to First Value > Create your workspace, test a support agent, trial a workflow, and connect an AI assistant to manage both through MCP. Start with one customer question and one measurable result. You do not need to configure every StateSet service before seeing value. **Fastest app path:** [sign up for ResponseCX](https://response.stateset.com/sign-up), open [Setup](https://response.stateset.com/setup), and create your first agent. Then add a source, test an answer, and decide where to launch it. **Provision multiple services together:** use [Brand Onboarding](https://onboarding.stateset.com). Preview and launch your selected services, then follow **Open your Response workspace** into Setup. If your organization already has a workspace, sign in to it rather than creating another. See the [portal walkthrough](/stateset-onboarding#the-portal--onboardingstatesetcom). ## Choose your first outcome Start in ResponseCX with one policy. Finish with an answer checked against your source. Connect Gorgias and replay a ticket. Inspect its draft and review or escalation decision. Verify the workspace through MCP, then authorize one bounded task. Create an order and check stock with the embedded engine. No hosted credentials needed. Not sure which service fits? Compare [Rust engines and MCP capabilities](/guides/rust-engines-and-mcp). ## What to have ready Keep your IDs and results in the [copyable onboarding record](#keep-a-record-you-can-resume) if you plan to continue across sessions or hand the task to an AI assistant. For the first agent, bring a work email, your company name, and one authoritative source such as your shipping FAQ or returns policy. Pick a colleague who can judge whether its answers are correct. For a helpdesk workflow, also have permission to connect your helpdesk and store. The accelerated trial below uses Gorgias, with Shopify when store context is needed. Without connector access, start with the agent walkthrough while your administrator connects the systems. For MCP, you need an MCP-capable assistant, its own model access, and an authorized connection to each StateSet server you plan to use. A browser login alone does not configure that connection. ## Signup to first answer Open [signup](https://response.stateset.com/sign-up) and complete the work-email sign-in flow. The default signup route leads into trial onboarding. Review the plan and terms shown; use your existing organization if your team has already joined. **Check:** you can open the intended organization, not just the sign-in screen. Continue to [Setup](https://response.stateset.com/setup). In **Choose agent**, select a use case and choose **Create my agent**. **Check:** the agent exists in your workspace. Keep its ID for tests and later handoffs. In **Test & launch**, choose **Add sources first** if the agent has no knowledge. Import one relevant policy and wait for processing before testing an answer. **Check:** the processed source contains the policy you want the agent to use. Run the [three-question walkthrough](/guides/first-support-agent#test-three-customer-questions). Test a known policy, missing context, and a policy boundary. Keep the conversation or evaluation result so your team can inspect it. **Check:** a reviewer can explain why each answer is acceptable or what needs correcting. Install the widget or connect a customer-facing channel only when the answers and escalation behavior match your policy. Follow the [channel setup](/guides/first-support-agent#5-choose-a-customer-facing-channel) and verify the installed or connected surface independently. **Check:** distinguish a successful test from a channel that customers can actually reach. **Test & launch uses a real chat runtime.** Treat its test messages as real generations, not a static mock. Workflow **shadow mode** is a separate setting in the automation engine; it does not make the Response widget a shadow-only experience. ## Measure time to value with outcomes Use these as milestones, not a guaranteed deployment schedule: | Milestone | Record | What it does not prove | | --------------------------- | -------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | Workspace ready | Organization and selected services accessible | Agent quality or connector readiness | | First useful answer | Test question, response, and policy used to judge it | Readiness for every support intent | | First workflow draft | Trial/run ID, draft, and review result | A reply was sent to a customer | | First reviewed live outcome | Approved dispatch and its run/conversation evidence | That human review can be removed | | Repeatable value | Useful-answer rate, review effort, escalations, and completed outcomes over a defined period | That a configured or activated agent is delivering value by itself | For connected workflow onboarding, aim for a **first real draft within a 30-minute onboarding session** when credentials and source material are ready. This is the engine team's onboarding target, not a service guarantee; imports, access approvals, and connector health can extend it. The workflow funnel reports actual progress through `first_event_received`, `first_draft`, and `first_reply_sent`. Use the [workflow trial](/guides/first-workflow-trial#see-where-onboarding-is-stuck) to find the next incomplete step rather than restarting setup. ## Keep a record you can resume Copy this worksheet into your project notes before starting. It is a suggested team record, not an API payload or a file the platform imports. Complete only the fields for your chosen path; leave unverified results as `unknown` and unused services as `not used`. ```text theme={null} FIRST OPERATION Outcome: [one question or operational task] Success means: [the observable result a reviewer will accept] Owner and reviewer: [names or team roles] Environment: [local / test workspace / production workspace] IDENTITY Response organization: [verified organization ID, or not used] Response agent: [agent UUID, or not created yet] Workstream brand: [brand UUID, or not used] Local database: [path, if using embedded commerce] PROGRESS Last completed step: [step and date/time with timezone] Evidence: [response, evaluation, job, or workflow run IDs] Observed result: [what actually happened] Expected behavior: [policy reference and what should happen] Current blocker: [unknown, missing access, processing, or failed check] Next action and owner: [one action and who can perform it] BOUNDARIES Permitted actions: [reads, specific creation, tests, or reviewed edits] Customer-facing surfaces: [widget, helpdesk, other channel, or none] Workflow routing and review gate: [verified values, unknown, or not used] Changes needing approval: [launch, routing, dispatch, or other actions] ``` Keep credentials in the connection UI or your credential store, not in this record. Reference restricted conversations by ID instead of copying customer details into a shared worksheet. A policy reference helps a reviewer judge the answer; a run ID helps them inspect execution. Neither substitutes for the other. ### Resume with your AI assistant Connect the appropriate [MCP server](/guides/manage-platform-with-mcp) before using this prompt. Paste the completed record after the prompt. The assistant must verify current state because configuration, access, and running jobs may have changed since your last session. ```text theme={null} Help me resume the StateSet onboarding record below. Start with reads only. Discover the tools and their schemas on my connected servers. Verify the workspace, agent, and brand identities against actual results. If the intended identity is missing or ambiguous, report that blocker before proposing writes. Check the last recorded result and any pending job or run. Do not create a replacement agent, restart an accelerator, or repeat a write merely because this is a new conversation. Separate saved configuration, queued work, completed tests, generated drafts, and confirmed customer-facing outcomes. Return: 1. What is verified now, with supporting IDs. 2. What is unknown, incomplete, or different from the saved record. 3. The next single action, its prerequisites, and whether it changes state. 4. An updated progress section for my record. Make no changes during this inspection. If a required tool or permission is missing, explain the missing capability rather than claiming verification. Onboarding record: [paste the completed worksheet] ``` The prompt guides behavior; server scopes and host approvals enforce permissions. For a failure, use [onboarding troubleshooting](/guides/onboarding-troubleshooting) before repeating setup. For a successful first result, continue to [your first week](/guides/first-week-with-stateset). ## Expand after the first result Follow [Your First Week with StateSet](/guides/first-week-with-stateset) for a launch worksheet, reviewed live outcomes, a correction-to-evaluation walkthrough, and a weekly value review. Add one intent or channel at a time. Turn a corrected answer into an evaluation, compare new configuration against those cases, and inspect the next real results. Use [MCP management](/guides/manage-platform-with-mcp) to let your assistant gather evidence, propose changes, and perform the operations you authorize. If a step stalls, use [onboarding troubleshooting](/guides/onboarding-troubleshooting) to identify the first unverified result. If access or deployment differs from these screens, capture the page, workspace/brand ID, and error without credentials and use the [support checklist](/support#report-an-api-problem). These walkthroughs were checked against the onboarding, Response, and Temporal source on 2026-09-20; availability still depends on the features enabled in your workspace.