Not sure which AI to send a captured task to? Each dispatch target in Notestream does a specific kind of work. Here's how to route your queue so the right AI handles the right job.
Your Notestream queue is a holding space for ideas, commitments, and deadlines, each classified on arrival and ready to move. The question isn't whether to dispatch. It's which AI gets the task.
The four main dispatch targets built into Notestream are in-app LLM chat (Claude, ChatGPT, or Gemini), Claude Code, Claude Cowork, and Lovable. They're not interchangeable. Each handles a specific mode of work well, and routing a task to the wrong target costs you more time than it saves. Getting the routing right means the AI returns something usable without you finishing it manually.
A Framework for Choosing the Right Target
Before you route, ask what the task requires as an output:
- Does it need to be written, researched, or reasoned through? Send it to in-app chat.
- Does it need to become working code? Send it to Claude Code.
- Does it need to run on a schedule, coordinate files, or repeat across steps? Send it to Claude Cowork.
- Does it need to be designed or prototyped visually? Send it to Lovable.
Each AI completes work in one mode. A writing task sent to Claude Code comes back as a function. A coding task sent to Lovable comes back as a layout. Matching the task type to the dispatch target is the core skill of working with a multi-AI stack, and Notestream's queue is where you develop that habit.
When to Dispatch to In-App Chat
In-app chat (Claude, ChatGPT, or Gemini) is the right first stop for cognitive work. Writing, analysis, research, editing, and planning all belong here.
You captured "Draft a positioning statement for the new pricing page." That's a writing task. Send it to Claude via in-app chat and work through the output in the conversation. You can refine the draft, ask follow-up questions, and iterate without leaving Notestream.
The same applies to exploratory tasks. You captured "Figure out how to structure the onboarding email sequence." That's a thinking task. Claude can draft a sequence structure, explain tradeoffs between approaches, and help you get to a clear direction before anyone writes a line of code.
One principle worth keeping in mind: in-app chat is best when the output is text you will read, revise, and act on yourself. It's conversational, iterative, and good at open-ended problems. When the output needs to run or be built, you're better off in Claude Code or Lovable.
Notestream's in-app chat lets you choose which LLM handles the conversation. Claude, ChatGPT, and Gemini each have strengths. The dispatch hub doesn't pick the model for you. You choose based on what you already trust for that kind of work.
When to Dispatch to Claude Code
Claude Code handles tasks that need to become working code, whether that's a new feature, a bug fix, or a refactor.
You captured mid-development: "Add a rate limit to the POST /captures endpoint, 10 requests per minute per user." That's a specific implementation request with enough context to act on. Claude Code can read your codebase, understand the existing auth patterns, and write the code without you describing the whole system from scratch.
Claude Code is also the right target for debugging tasks. If you captured "Investigate why the classification step is timing out on long inputs," that's something Claude Code can dig into directly, looking at your actual implementation rather than describing a general approach.
One useful pattern: capture the task the moment it surfaces, then dispatch it when you're in build mode. A requirement that comes up during a client call, a Claude conversation, or a code review lands in Notestream immediately, worded the way you said it, classified as a task. When you're ready to build, the context is still intact.
When to Dispatch to Claude Cowork
Claude Cowork handles tasks that repeat, coordinate files, or span multiple steps.
You captured "Every Monday, pull last week's blog metrics and send a summary to the team Slack channel." That's not a one-time writing task. It's a recurring automation. Claude Cowork can build and run it on schedule, without requiring your attention each week.
Cowork is also the right target for file-heavy tasks. Reorganizing a folder structure, generating a weekly report from a set of source files, running the same transformation across multiple documents. Any task that would take several manual tool calls to complete is a candidate for Cowork.
A simple routing rule: if the task description ends with "...and then do that again every week," it belongs in Claude Cowork. If it's a one-time task that touches multiple files or steps, Cowork is still likely the better choice over in-app chat.
When to Dispatch to Lovable
Lovable handles tasks that need visual output: layouts, prototypes, and UI components.
You captured "Redesign the settings screen: move the notification toggles to the top, add a section for billing history." That brief has enough specificity for Lovable to generate a working layout. Send it to Lovable and you get a prototype you can review, revise, or ship without writing a design spec or opening Figma first.
Lovable also works at the exploratory stage. You captured "What would a calendar view in the task list look like?" That's open-ended design exploration. Lovable can show you a direction in minutes. You review it, update the brief, and iterate without involving a designer until the direction is clear.
The distinction from Claude Code: Lovable produces visual components. Claude Code produces working logic. For tasks that need both, the typical path is Lovable first for design, then Claude Code for implementation.
Putting the Routing Decision Together
When you open your Notestream queue, you're looking at classified items ready to move. Here is a quick example of a real morning queue and how each item gets routed:
"Revise the About page intro to emphasize the dispatch feature." Routes to in-app chat with Claude.
"Fix the bug where email forwarding fails on addresses with plus signs." Routes to Claude Code.
"Pull this week's published blog posts and format them into a weekly recap for Slack, every Friday at 9 AM." Routes to Claude Cowork.
"Mockup a new onboarding screen for the mobile app." Routes to Lovable.
Same queue. Four tasks. Four dispatch targets. Each one goes to the AI that can complete it without you bridging the gap.
Capture Ideas From Anywhere, Dispatch to the Right AI
The routing skill develops quickly once you start treating Notestream as the central hub for your AI workflow.
Ideas, commitments, and deadlines arrive in the AI assistant field, via forwarded email, or through the MCP API from tools like Claude Code or Cursor. Each capture is classified on arrival: task, reference note, sub-task, or reminder. By the time you open the queue, the item is already sorted and named clearly enough to dispatch.
The routing question then reduces to one decision: which AI gets this task right now? In-app chat for thinking. Claude Code for building. Claude Cowork for repeating. Lovable for designing. Make that call, dispatch, and move on.
You're not just organizing notes. You're routing work to the right AI at the right moment. The queue handles the holding. The dispatch handles the handoff.
Try it free at notestream.ai.