My Dream AI Client
After extensive use of various AI clients, I’ve identified three key features needed for an ideal platform.
Feature 1: Universal Chat
Building on established demand, this capability should include:
- Model selection flexibility
- Support for numerous MCP servers and tools without arbitrary limits
- Cross-client memory storage and retrieval
- Integration with knowledge platforms like Notion and Confluence for searchable references
- Workflow triggering via tagging syntax (e.g.,
@post-meeting-agent)
StackChat is a leading example due to robust remote MCP support and multi-model selection.
Feature 2: Workflows
Described as “a sequence of steps I regularly do,” workflows handle repeatable tasks like meeting followups and documentation updates. Ideal implementation requires:
Structure:
- Bucket organization similar to projects
- System prompts guiding workflow behavior
- Defined tool lists with core/secondary designations
- Accessible resource libraries (documents, folders, codebases)
- Per-workflow model selection
Execution:
- Schedule, webhook, or chat-based triggering
- Feedback mechanisms updating prompts and storing successful outputs
Relevance AI approaches this capability, though tool portability limitations present challenges.
Feature 3: Initiation Points
Multiple access methods enhance usability:
- Desktop application as primary interface
- Browser extension functionality
- Raycast integration
- Slack thread tagging
- Resource pooling across systems