Talkable AI#
Talkable AI is the agent layer of your referral program. Use the AI Assistant in your Talkable dashboard, or connect Claude Code, Cursor, Codex, VS Code, or any MCP-compatible agent and work with Talkable from there. Either way, you ask in plain language and the agent works through the same data, reports, and campaign settings you use in the dashboard: it explains performance, resolves support requests, builds and optimizes campaigns and Wallet passes, and shows you every change to your program before it is applied.
Agent connections are account-isolated, role-based, and revocable. The agent can do only what your role or your key allows, campaigns are built and changed in test mode with a preview of every change, nothing goes live until you launch, and eligibility, fraud checks, and reward issuance are decided by Talkable’s deterministic logic, never by the model.
Note
Talkable AI is enabled per site by your Customer Success Manager. Capabilities marked Early access are available to selected customers today; ask your CSM to join.
Where you use it#
In the Talkable dashboard. Open a site and choose AI Assistant. Conversations are scoped to the site and saved, answers about how Talkable works cite the documentation, and every proposed change comes with a preview you approve.
In your AI agent. Add Talkable as an MCP server in Claude Code, Cursor, Codex, VS Code, or any MCP-compatible client: one URL and a sign-in. Your agent gets every capability on this page under your permissions, across every site you can access, or delegates a whole task to the Talkable assistant and gets the result back. For scripts and automations, create an API key scoped to a site or account instead of signing in. See Connect your agent.
Capabilities#
Know your program#
Performance analysis. Any metric your dashboard shows, filtered and segmented by campaign, period, device, traffic source, or A/B test variant, with the numbers separated from the explanation.
Reports and exports. The right report type for the question, a scheduled CSV export, and the download when it is ready.
Campaign and settings visibility. How each campaign is set up, where it shows, what rewards it uses, the current site and account settings, and who changed what in the settings history.
Purchase tracking. Whether the Talkable script is live, purchases are being tracked, and your e-commerce platform integration is connected, so your reports are complete.
Product knowledge. How Talkable works, answered from the product documentation and your site’s configuration, with citations.
For example: “Referral revenue grew in June versus May. Show which campaigns and traffic sources drove it, and how share rate and friend conversion changed.”
Resolve support requests#
Referrals, people, and coupons. Look up a referral, a person, or a coupon, understand what happened, and act within your role: approve or void a referral, unsubscribe, resubscribe, or anonymize a person, expire a coupon code. Each action runs with your confirmation. Your support team can do all of this with Talkable’s dedicated Customer Support role. See Referral approval and the Customer Service Portal.
For example: “Look up jane@example.com, show her referrals and rewards, and draft a reply to her question about her reward.”
Build and optimize campaigns (Early access)#
Create campaigns. A new campaign from a preset or a clone, in test mode, with the rewards, placements, audience, and schedule you describe.
Incentives and rewards. Add or change incentives: reward type, amount, and qualification criteria such as a minimum order value.
Placements. Where a campaign shows: device, URL rules, appearance, and the triggering event. See Placements.
Copy and configuration. Any copy or configuration value in the Editor, changed in one batch with a before-and-after preview.
Templates, emails, and CSS. A view’s HTML and CSS edited with a syntax check and a colored diff, new views such as an extra reminder email, and reusable template fragments.
Images and assets. An image generated from a description or edited from one you attach, and files you attach uploaded to the site’s asset library and placed in the campaign.
A/B testing, end to end. The agent proposes variants based on your campaign’s copy and goals, sets up the test on the copy you choose, tracks it by variant, and concludes it with the winner.
Coupon lists. A new coupon list, or single-use codes appended from a file you attach.
Test purchases and launch readiness. A test purchase to see the campaign as a customer, then a readiness check and the link to the launch form.
Site settings and brand. Site-wide settings and your brand profile, such as logo, colors, fonts, and brand voice used across campaigns, changed with a preview.
For example: “Clone the Spring Sale campaign as Summer Sale, set a $25 advocate reward for orders over $100 and a 15% friend coupon, replace the hero image with the one I attached, and show me the preview.”
Build Wallet passes (Early access)#
A campaign can carry an Apple Wallet and a Google Wallet pass with the customer’s referral link, rewards, and offers. See Talkable Wallet. The agent configures the pass with the same tools it uses for the rest of the campaign.
Pass design. Colors, images, the fields on the front and back of the pass, links, and the share message, in your brand.
Where the pass is offered. The places where customers are offered the pass: the share page, the referral dashboard, the claim page, reward emails, or a standalone install page.
Push notifications. The welcome message, share and reward reminders, and referral updates the pass sends, with their text and timing. See Push notifications.
Publishing. The updated design pushed to every installed pass.
Pass reporting. Installs, taps on the pass, sales from push notifications, and referrals shared from the pass.
For example: “Update the Refer a Friend campaign’s Wallet pass with the new logo and brand colors, add a welcome notification and a share reminder after three days, and show me the preview before publishing.”
Integrations and webhooks#
App Store integrations. Integrations from the Talkable App Store, such as your email service provider or reward and shipping providers: what is installed and what is available, an integration installed, configured, and activated, and a test run before it goes live.
Custom App webhooks. Campaign events sent to your own systems through Custom App, with the trigger and payload you choose, and a test payload delivered to your endpoint.
For example: “Install the Custom App, send referral approved events to https://hooks.example.com/talkable, and send a test payload.”
Manage users#
Users and roles. Users invited, site roles set, site access removed, password resets sent, and account admins managed.
For example: “Invite maria@example.com with the Customer Support role on all our sites, and make john@example.com an admin of this site.”
Work your way#
Skills. A multi-step task saved as a skill and run as one command, for example a weekly funnel review or a campaign launch checklist, shared with your team on the site. Running skills is available to every role; saving new ones is in early access.
Memory (Early access). Facts and preferences you save about a site, such as naming conventions or reporting periods, that the agent uses in later conversations.
Your agent, your workflow. From Claude Code, Cursor, or your own automation, the same tools across every site you can access, or a task delegated to the Talkable assistant with the result returned to your agent.
For example: “Compare friend conversion rate for the last 30 days across all my sites and rank them.”
You stay in control#
Your permissions, your scopes. For a user, the agent can do only what that user’s role allows on the site. For a script or automation, an API key you create with the scope you choose, read-only or read-write, for one site or the whole account.
Preview, then apply. Every change to your program is shown before it happens. You approve it, and the agent applies exactly what you saw, nothing else.
Build in test mode, launch when ready. The agent builds and edits campaigns in test mode, where you can review them exactly as customers will and run test purchases. Launching is your decision.
Logged. Every agent action is recorded with who asked, which tool ran, when, the outcome, and the source, including previews that were not applied. Program changes appear in Settings Changes under your name.
Deterministic decisions. Eligibility, fraud checks, and reward issuance are decided by Talkable’s deterministic logic, the same code that runs your program today, never by the model.
An agent connected over MCP with a write credential applies changes immediately, without the preview. Grant that credential deliberately; you can revoke it at any time. See Permissions and safety.
Connect your agent#
Any MCP-compatible client connects to Talkable with OAuth, a one-time sign-in in your browser, or a named API key. In Claude Code:
claude mcp add --transport http Talkable https://www.talkable.com/mcp
Cursor, Codex, VS Code, and other clients use the same URL in their MCP settings. One connection covers every site your credential can access. A read-only credential never sees write tools. See MCP Server for setup, keys, and scopes.
Learn more#
Create, edit, test, translate, and prepare campaigns and Wallet passes for launch |
|
Metrics, exports, purchase tracking, and support actions by role |
|
Roles, credentials, preview and approval, logging, data handling |
|
Connect Claude Code, Cursor, Codex, VS Code, or a custom agent |
Contact us
Interested in setting this up? Contact your CSM or get in touch here.