ForwardAgentKit

Embed AI agents in native Apple apps, connected to your models, tools and data.

Core capabilities

Agent runtime
Streaming replies, tool execution, cancellation and queued follow-ups. Save and restore conversation state, including app context.
Swift tools & app context
Expose your app’s actions as Swift functions with JSON Schema inputs. Supply reference data and scope file operations to authorized workspaces.
Models & AI capabilities
Connect supported providers or custom OpenAI-compatible HTTPS endpoints. Call text, image, audio, music, video, search, extraction, embedding and decision APIs directly or through Agent tools.
Native conversation UI
Add optional chat and AI settings components, with streaming Markdown, attachments, tool activity and model selection. Or build your own UI on the session API.
Apple system services
Opt into native calendar, reminders, contacts, files, media and Mac services. Access follows system permissions, app configuration and approval callbacks.
Skills & MCP
Load reusable Skill instructions and connect remote MCP servers over Streamable HTTP. Your app controls which resources and tools each Agent can use.
Agent memory (in development)
Retain preferences and project context across conversations, with per-Agent enablement and shared or private memory. Automatically learn from conversations; inspect, correct or forget saved memories. Full-text retrieval needs no embedding model. Optional local or cloud embeddings add semantic search.

Use the pieces you need

Link the runtime, then add UI, system services or extensions as separate SwiftPM products. Your app owns credentials, data, permissions and persistence.

AI features depend on the connected provider. Native services depend on the OS, device and granted access.

Built with ForwardAgentKit

Package access

Currently distributed as a private Swift package.

Discuss an integration