The product
Sidenet gives vertical SaaS companies the full stack — build, deploy, observe, and monetise AI assistants without building the plumbing.
Embedded with 1 line of code
Configurable with code or no code
Build
Everything you need to go from zero to a working multi-agent system — models, memory, tools, and orchestration — configured visually in Studio.
Build agent networks visually in Studio. One orchestrator reads every request and routes to the right specialist — or chains several together. Each agent gets its own prompt, tools, guardrails, and model.
Guest ops invokes scheduling. Scheduling calls the calendar API. The calendar checks availability and books a slot. Your users see one clean answer — the orchestrator manages the chain behind it.
"Book a spa appointment for the guest in 412"
Routing to Guest Ops Agent
Delegating to Scheduling Sub-Agent
Calling Calendar API tool
Available: 2:00 PM, 3:30 PM, 5:00 PM
"Booked: Spa at 2:00 PM for Sarah Chen, Room 412"
The orchestrator routes dynamically based on intent, context, and the agents available — not hardcoded rules. No YAML, no decision trees, no rewiring.
Four memory modes — all configured in Studio, no code after setup.
Builds a working picture of each user — preferences, projects, context — and carries it forward between sessions.
Tracks what's been said and decided as conversations grow. Never loses the thread.
Retrieves actual excerpts from past conversations when something relevant comes up.
Control how many recent messages stay in view. Dial up for complexity, keep lean for simple Q&A.
OpenAI, Anthropic, Google, Mistral, Meta, Cohere — 60+ providers. Best model for each agent. Swap in two clicks. New models within 24 hours of release.
60+ providers supported
Connect 1,000+ external apps, your own internal APIs, or both. Platform credentials for internal services, per-user OAuth for their calendar or CRM — mix both in the same network.
One API key for all users
Per-user OAuth tokens
Paste an OpenAPI spec and every endpoint becomes an agent tool in seconds. No manual mapping, no SDK wrapping. Internal services become callable instantly.
Deploy & Embed
Two ways to access your agents — embed the Chat SDK directly in your product, or call them via API. Three isolated environments and instant rollback keep your customers safe.
A branded, context-aware chat appears inside your product. It sees the page, the data, and the user's role — no extra context needed. Customise colours, fonts, icons, and welcome messages from Studio. Light and dark mode built in. Responsive, keyboard-navigable, feels native.
This isn't a generic chat window. The copilot sees the page, the data, and the user's role. A front desk agent asking about a guest doesn't need to specify which reservation — the copilot already knows.
What's the guest's room preference?
Sarah Chen (Room 412) prefers king bed, extra pillows, and late checkout.
Hit your agents and agent networks directly through the Sidenet API. Same orchestration, same tools, same memory — no frontend. Build custom interfaces, trigger agents from backends, or wire into existing workflows.
Development, staging, production — fully isolated. Test in Studio, preview through the SDK in staging, promote when ready. Something break? Roll back in seconds, not sprints.
Build and test agents in Studio. Full traces, logs, and debugging. Break things, tune prompts, and experiment safely.
Preview your network through the SDK in your staging environment — exactly what customers will see, before they see it.
Promote to production when you're ready. Roll back to the previous version in seconds if anything unexpected happens.
Each environment is fully isolated — no shared state or config
Observe & Iterate
Full observability across every trace, thread, and tool call. Scoring catches regressions. User feedback closes the loop. Your assistants get measurably better every week.
Each interaction generates a complete trace — orchestrator routing, agent execution, tool calls, response. Drill into any span for timing, tokens, cost, and exact I/O. When something breaks, you know exactly where and why.
See every conversation for any user, across sessions. Spot patterns, track resolution rates, review how context builds over time. Thumbs up/down and written feedback are collected directly in the copilot — filter by agent, time, or sentiment in Studio.
Code-based scorers for hard rules — valid JSON? Right tool called? LLM-based scorers for subjective quality — helpful? Accurate? On-brand? Runs continuously on a sample. Scores drop, you get flagged.
Compare prompt versions and model swaps side by side — v1 scored 4.1, v2 scored 4.5. Ship v2. Done.
Track & Monetise
Full cost ledger per customer, per assistant, per model. Simulate pricing before you ship. Monitor margins in real time.
Cost per customer, per assistant, per model, per tool call. No guessing whether AI features are profitable. No month-end surprises.
Model usage-based, tiered, per-seat, or hybrid pricing. See how each strategy affects margins at different usage levels. Commit with data, not assumptions.
Revenue vs. cost, tracked live. Alerts when a customer's usage erodes your margin. Know which assistants drive profit and which drag it down.
Get started
One SDK call. A branded, context-aware copilot appears inside your app. Production-ready from the first line.
Create an orchestrator and specialist agents in Studio. Assign tools, prompts, memory, and sub-agents — visually.
Three environments. Test internally, preview in staging, promote to production. Roll back in seconds.
Each environment is fully isolated — no shared state or config
Traces, feedback, scoring, usage analytics. Your assistants get measurably better every week.
Embed the copilot. Build your agents. Iterate in Studio.