Yedric
ShopPad · Yedric: Embeddable In-App AI Agent That Takes Actions Inside Your SaaS or Shopify App
Yedric is an AI assistant that software companies embed in their SaaS or Shopify app with one script tag, so users can ask for a task and have it carried out through the app's APIs or MCP tools. It suits product and engineering teams that want an in-product agent without building one, and it is currently free to start, with your own OpenAI, Anthropic, or Google keys for other models.
Best for
SaaS companies and Shopify app developers who want an in-product assistant that performs actions through their own APIs or MCP tools, and who are comfortable bringing their own model keys
Not ideal for
Teams without engineering capacity to expose APIs as actions, companies that need self-hosting or documented compliance certifications, and businesses looking for a full help desk or an agent for their storefront's shoppers
Who it's for
Product and engineering teams at SaaS companies and Shopify app developers who want to add a natural-language, action-taking assistant to their own product
Capabilities
- One script tag loads a <yedric-widget> web component that renders in Shadow DOM, with no build step or other dependencies
- Custom actions from a connected MCP server, an imported OpenAPI spec, tools generated by crawling your API docs, or hand-written definitions
- Per-action approval: Automatic, Ask first (the user confirms before it runs), or Off
- Client-side tools registered with window.Yedric.registerClientTool run in the host page, for example to open a modal or read page state
- Built-in tools for knowledge base search, multiple-choice clarifying questions, tables, Mermaid diagrams, navigating the user to a page, and handing off to a human
- Knowledge base from a crawled website (re-scraped weekly), PDF, Word, or text uploads, and written content
- Page prompts that change suggestions, override instructions, and restrict which MCP servers are active based on the current URL
- Secure mode ties sessions to a verified user with an HMAC-SHA256 signed username, with Shopify App Bridge session tokens supported for embedded Shopify apps
- Dashboard with conversation volume, sentiment, frustration tagging, feature requests, and ranked recommendations for prompt edits and documentation gaps
- Simulations with saved scenarios, suites, pass/fail criteria, and scheduled regression runs, plus version history with diff, restore, optional GitHub sync, and JSON import and export
- Built-in live chat inbox for human handoff in the same widget, or an external handoff event for your existing support tool
- Hosted MCP server (OAuth) for managing assistants and reading conversations from Claude Desktop, Claude Code, Cursor, or ChatGPT, plus outbound webhooks
Limitations
- Only the default GPT-4o Mini model runs without a key; every other model needs your own OpenAI, Anthropic, or Google key, and those usage costs are billed by the provider
- Actions beyond the built-in tools need developer work: you have to expose your app's operations as an MCP server or API and handle authentication for them
- No self-hosted deployment is documented; the dashboard, chat backend, and knowledge base run on Yedric's hosted service
- Pricing for large-scale use is not published: the docs say to bring your own keys for advanced models or scale and to contact the company for details, and plans are subject to a fair use policy
- Without secure mode, sessions are anonymous and browser-local, so per-user history and username-based reports require server-side signing
- No compliance certifications such as SOC 2 are documented, and no native mobile SDK is documented; the widget is a web component
- A young product: the public changelog starts in May 2026 and the public launch was in October 2026
Use cases
- Letting users of a SaaS app ask for an export, a settings change, or a fix to a disconnected integration and having the assistant run it through the app's API
- Adding an in-app assistant to an embedded Shopify app, authenticated with Shopify App Bridge session tokens
- Answering product questions from a crawled help center while handing unresolved conversations to a human in built-in live chat
- Mining assistant conversations for feature requests, pain points, and documentation gaps to feed the product roadmap
- Showing page-specific suggestions, such as order questions on an orders page, beside the closed chat launcher
- Regression-testing prompt and tool changes with saved simulations before they reach users
Our take
Yedric is a good fit for SaaS and Shopify app teams that already have a usable API or are willing to build an MCP server, because the action layer is where it differs from a help widget, and the per-action approval modes, secure mode, and simulations make that safer to roll out. The free-to-start, bring-your-own-key model keeps costs predictable today, but the terms for large-scale use are not published, and there is no self-hosting or documented compliance certification yet. Start with knowledge base answers and a few low-risk actions set to ask first, then use the dashboard's feature request and documentation-gap reports to decide what to expose next.
Who should use it
SaaS companies and Shopify app developers with an API or MCP server who want users to get tasks done by asking, product teams that want conversation-derived feature requests, and small teams that want an in-app assistant without paying per seat or per resolution.
Who should skip it
Teams with no engineering capacity to expose actions, organizations that need self-hosting or formal compliance attestations, support teams that need a full ticketing help desk, and merchants looking for an AI agent for their own storefront customers.
Strengths
- Takes actions through your own APIs and MCP tools, with per-action approval
- Currently free to start, with no markup when you use your own model keys
- One script tag to embed, with secure per-user sessions
- Turns conversations into feature requests, pain points, and documentation gaps
- Simulations, version history, and live chat handoff included
Weaknesses
- Useful actions need engineering work on your APIs
- Only one default model works without your own key
- No self-hosting or documented compliance certifications
- Young product with unpublished terms for large-scale use
Yedric pricing
Free
$0
- Unlimited assistants
- Unlimited conversations
- Fully customizable widget
- Bring your own OpenAI, Anthropic, or Google keys with no markup
Free tier limits: New assistants run on GPT-4o Mini using Yedric's keys; any other model requires your own provider key. Each knowledge base has a storage limit, and plans are subject to a fair use policy.
Note: Yedric does not publish paid plans. Its docs say it is completely free to get started and that more advanced models or large-scale use require your own Anthropic, OpenAI, or Google keys, with the company asking large users to contact it for details. Model usage on your own keys is billed by the provider.
Technical specs
Modalities
Text, File uploads
Available models
Where Yedric excels
Natural-language control of an existing SaaS app
Exposing operations such as exports, settings changes, or integration repairs as actions lets users ask for the outcome, and Yedric runs the matching API call, with Ask first approval for anything sensitive.
Support for an embedded Shopify app
With Shopify App Bridge authentication enabled on the tool server, the widget authenticates tool calls as the current shop without a hand-managed token, and an optional built-in tool can ask merchants to review the app on the Shopify App Store.
Product discovery from conversations
The analyzer tags feature requests, pain points, and frustration and the dashboard ranks documentation gaps and prompt fixes, each linked to the conversations behind it, which gives product teams a steady signal about what users cannot do yet.
Yedric vs. competitors
Yedric vs. Chatbase
Chatbase is a self-serve builder for AI support agents trained on your content and deployed to a website, WhatsApp, email, Slack, and help desks, with actions such as custom API calls and plans priced by monthly message credits. Yedric is currently free to start, with your own model keys for other models, and is built mainly as an in-product layer for a SaaS or Shopify app, with MCP and OpenAPI actions, client-side tools, signed user sessions, and page-specific behavior. Chatbase fits teams that want multi-channel support coverage, and Yedric fits product teams that want an assistant acting inside their own app.
Yedric vs. Fin
Fin resolves customer questions inside Intercom or another help desk from help center content and can take actions through APIs, Data Connectors, and MCP, at $0.99 per outcome. Yedric is not tied to a help desk: it is embedded directly in your product, uses your own model keys at no charge from Yedric, and pairs actions with product analytics such as feature-request detection. Fin fits support teams measuring ticket resolution, and Yedric fits SaaS teams that want users to complete tasks in the app by asking.
Frequently asked questions
What is Yedric?
Yedric is an AI assistant that a software company embeds in its own SaaS or Shopify app. End users ask for what they want in natural language, and Yedric answers from your knowledge base or performs the task through actions you expose, such as your API or an MCP server. It also analyzes conversations to surface feature requests and documentation gaps for your team.
How much does Yedric cost?
Yedric says it is completely free to get started, with unlimited assistants and conversations. New assistants run on GPT-4o Mini using Yedric's keys. For other models or large-scale use you add your own OpenAI, Anthropic, or Google keys, which Yedric says carry no markup, and the docs ask large users to contact the company for details.
How does Yedric take actions in my app?
You add actions on the Actions tab by connecting an MCP server by URL, importing an OpenAPI spec, letting Yedric generate tools from your API docs, or writing definitions by hand. Tools are normally called server to server, and you can also register client-side handlers in the page. Each action can run automatically, ask the user first, or be turned off.
Which AI models does Yedric support?
Models from OpenAI, Anthropic, and Google. GPT-4o Mini works with no key; any other model needs the matching provider key added under AI Service Keys, and the model dropdown shows only models whose provider key is set.
How does Yedric handle user identity and security?
In secure mode, your server signs the user's identifier with an HMAC-SHA256 secret and the widget sends the signature, so sessions and history are scoped to a verified user. Embedded Shopify apps can use Shopify App Bridge session tokens for tool calls. Widget keys can be limited to allowed domains, and secrets for your tools are stored encrypted.
Can Yedric be self-hosted?
No self-hosted option is documented. Yedric runs as a hosted service, and you embed its widget in your app.
Yedric vs Intercom Fin: what is the difference?
Fin is a support agent that resolves customer questions inside Intercom or another help desk and is priced per outcome. Yedric is an embeddable in-app assistant, currently free to start, built mainly to carry out tasks in your own product through actions you define, using your own model keys, with conversation analytics aimed at product teams.
Integrations & fit
Alternatives to consider
Chatbase
Small teams that want a self-serve AI support agent on their website or in their existing help desk, with published monthly plans
FreemiumSalesforce (Fin, formerly Intercom)
Support teams on Intercom, Zendesk, Salesforce, HubSpot, or Freshdesk who want to automate resolution of repetitive questions at a transparent $0.99-per-outcome price
PaidAbout Yedric
Yedric is built by ShopPad, a Shopify app maker that first used it inside its own apps, and is aimed at the companies that make software rather than at their end users. You create an assistant in the Yedric dashboard, give it instructions and a knowledge base (a crawled help site that is re-scraped weekly, PDF, Word, or text files, or written notes), and paste a script tag that loads a web component into your app. What separates it from a help widget is the Actions layer: you connect a custom MCP server by URL, import an OpenAPI spec, point Yedric at API documentation so it generates the tools, or define actions by hand, and you can also register handlers that run in the host page to open a modal or read page state. Each action is set to run automatically, ask the user first, or stay off, and page prompts can change suggestions, instructions, and the available tools per URL. Secure mode binds each session to a signed user identity (HMAC-signed usernames or Shopify App Bridge session tokens), which also enables per-user history and usage reporting. The second half of the product is for the vendor: a dashboard that tags sentiment, frustration, topics, and feature requests from conversations and proposes prompt edits and documentation gaps, plus simulations for regression testing, version history with restore and optional GitHub sync, built-in live chat handoff, proactive alerts, webhooks, and a hosted MCP server for managing assistants from Claude, Cursor, or ChatGPT. Yedric says it is completely free to get started: new assistants run on GPT-4o Mini on Yedric's keys, and other models use your own provider keys with no markup. The tradeoffs are that useful actions still need engineering work to expose and authenticate your APIs, there is no self-hosted option or published compliance certification, pricing for large-scale use is handled by contacting the company, and the product is young, with a public launch in October 2026.
Updates from Yedric
Yedric launched publicly on October 1, 2026, with a Show HN post from its maker and a Product Hunt launch where it was Launch of the Day, ranked #2 for the day.
Built-in live chat became available to all accounts, so teams can chat with visitors from the Yedric dashboard and offer a human handoff in the widget without asking Yedric to enable it first.
A new navigation tool lets the assistant open the relevant doc or screen instead of describing where to find it, and a Re-open when page changes setting keeps the same conversation open as users move between pages.
Are you the founder? Claim this listing →