Connect agents to the app you already have. Core brings execution, knowledge, workflows, and embedded UI together, so you can focus on the experience your customers use.
Define the work, connect your systems, and put the result in front of people. The pieces work together from the start.
Configure the goal, model, tools, and knowledge. Core handles model calls, context, tool use, and refinement, with reusable patterns for research, creation, and specialist collaboration.
See a team of agents at workStart from events or schedules. Stream progress to your app, wait for external signals, and deliver notifications through signed webhooks or your AWS SQS queue.
Create collections around your products, customers, or processes. Core manages indexing, embeddings, and search so agents can find relevant information and reference its sources.
Watch data become knowledgeFollow parallel tasks, see why a branch was chosen, and pick up where external work left off. Keep progress, results, and usage in view.
Bring your provider credentials. Choose a supported model, reasoning level, and budget for each agent, with AI provider charges kept separate from platform usage.
Core provides the agent infrastructure. Your team connects it to your product’s rules, systems, and users.
| Core handles | Your team supplies |
|---|---|
| Agent execution & workflow coordination Run agents, coordinate steps, and resume work after input or approval. | Domain logic & business tools Define objectives, implement business tools, and choose which actions need review. |
| Progress, input, approval & result interfaces React hooks and components show progress, collect responses, and present results. | The surrounding product experience Design your pages, style the components, and decide where agents fit into the user journey. |
| Knowledge indexing & retrieval Index supplied content and retrieve relevant material within configured access. | Data selection & access policy Choose source material, keep it current, and configure who and which agents may use it. |
| Integration mechanisms & scoped context Provide the server SDK, scoped client sessions, and events; enforce Core’s configured permissions. | Authentication & business-system authorization Authenticate your users, establish their scope, and enforce permissions in your business APIs. |
Your server creates a scoped client session after sign-in. Your Core API key stays on the server.
Add agentic capabilities to your product with an execution engine already built in. Configure agents and workflows, start work through an API, and receive progress and results in your app.
Set the objective and instructions, choose a model, and connect tools and knowledge. The platform manages model calls, tool execution, and the steps needed to produce the result.
Route progress and completion through signed webhooks or your AWS SQS queue, with retries and delivery history.
Call your tools, wait for an external task, then resume with its result. Your app can follow every step and fetch the output.
import { Jwtson } from '@jwtson/sdk';// Reads JWTSON_API_URL and JWTSON_TOKENconst jwtson = new Jwtson();// Creates it, publishes a change, or does nothingconst agent = await jwtson.agents.ensure({ name: 'Summarizer', instructions: 'Summarize the text in one sentence.',});const run = await agent.start('A long article...');for await (const text of run.textStream()) { process.stdout.write(text);}No. Agentic processing is built into the platform. You configure objectives, instructions, models, tools, and knowledge. The platform provides the execution loop, context handling, and reusable patterns for research, creation, review, and collaboration. Your team connects the business systems and defines the work, without building agentic processing from scratch.
Research that brings together several specialists. Onboarding that adapts to each customer. Campaigns that improve through feedback. Document studio turns connected knowledge into a finished document. Connect agents, tools, and everyday business steps into a workflow built around the outcome you want.
Yes. Document studio shows the whole process: search your knowledge by meaning with embeddings, gather the relevant material, plan the structure, and write a draft. Edit or refine the document, then save the version you want. Your application can receive the completion event and retrieve the finished work.
Yes. Run specialists in parallel, combine their results, choose a branch based on new information, or refine an output through a bounded feedback loop. Work can also wait for input or an external event and continue from that point. Use agents where reasoning helps and ordinary steps where the path is clear.
Start workflows through APIs, incoming events, or schedules. Connect tools to read data and take action in your systems. Stream progress into your interface and route notifications through signed webhooks or your AWS SQS queue. Your own integrations can use those events to update a CRM, send an email, or trigger the next process.
Yes. React components and headless hooks bring agent output, live progress, input and approval forms, and finished results into your product. Compose the experience around your design system. Your backend authenticates users and grants scoped access, so people can work with agents inside your app without a separate platform login.
Yes. Bring your provider accounts and choose a supported model, reasoning level, and budget for each agent. A research specialist and a writing agent can use different configurations. AI provider charges remain separate from platform usage.
Where you want them. Let a workflow run from start to finish, ask someone for missing information, or add a review before a particular action. Input and review are building blocks you choose for the work, and a waiting workflow can continue when the response arrives.
It fits behind your existing product. Your application keeps its interface and business records while agents handle multi-step work through your tools. APIs, progress streams, and events connect the experience, so your users can stay in the app they already know.
Start with one useful agent. Give your users a new way to get work done.