Build AI into your SaaS
Ship AI features without building the AI platform first.
Codapult gives your product one production path for chat, agents, tools, retrieval, guardrails, model routing and usage-aware operations.
One gateway · Provider control · Production metering
ai-request.trace
One execution path
Request
One entry point for every AI feature
Chat, agents and background jobs start from the same typed gateway instead of separate provider calls scattered through the app.
$ streamText() · generateText() · batch()
One execution path
Every AI workflow gets the same production controls.
Start with a chat response. Grow into agents, tools, retrieval and asynchronous work without creating a second AI architecture for every feature.
Chat and streaming
Ship responsive AI experiences with persisted conversations and one gateway.
Agents and tools
Let models call product logic through typed, permission-aware tools.
RAG and embeddings
Bring tenant-scoped knowledge into requests when retrieval is enabled.
Guardrails
Keep validation, authorization and policy checks inside every execution path.
Metering and budgets
See tokens, cost, latency and organization usage before surprises reach production.
Playground and batch
Compare prompts and models, then move repeatable work into asynchronous jobs.
AI inside the product
From first prompt to a feature your customers can trust.
The dashboard gives your team a real surface to configure prompts, agents, models, guardrails and usage — while the gateway keeps runtime behavior consistent.
Model catalogue with explicit IDs
Organization-aware quotas and metering
Playground before production release
Codapult workspacecodapult.app
Safe context, deliberate defaults
RAG is powerful because it is not automatic.
Enable retrieval when your product is ready. Index by organization, enforce admin permissions and keep embedding cost visible instead of turning context into a blind spot.
Opt-in
RAG stays off until you configure the providers and enable it.
Tenant-scoped
Documents and search results stay inside the active organization.
Measurable
Limits for documents, chunks, scores and batches remain explicit.
1
gateway
Chat, agents, batch and playground share the same execution contract.
5
control surfaces
Core, chat, agents, batch and playground can be enabled independently.
0
provider lock-in
Model catalogue and adapters keep provider decisions at the edge.
Build AI into a system you still control.
Inspect the gateway, flags and implementation in the docs.
Works with the model providers and coding agents your team already uses.
Next.js 16
TypeScript
PostgreSQL
Turso
Drizzle ORM
Tailwind CSS
shadcn/ui
Better Auth
Resend
tRPC v11
TanStack
Stripe
Lemon Squeezy
Vercel
Claude Code
Cursor
Codex