Pick the right AI work form for the job instead of forcing every task into a chatbot. Five agent types — Agent, FlowAgent, LoopAgent, AudioAgent and Workflow — extended flexibly through a powerful API.
Strong connectors, a knowledge RAG foundation and a model gateway — with MCP, APIs, skills and a secure sandbox to build AI into the business as fast as possible and keep it running reliably.
LogTree traces every message across 57 node types. Versions can be restored, and Insights tracks usage. GPTBots forward-deployed engineers co-build the first agents with your business team.
Build agents that search knowledge, query data and call tools on their own, from knowledge-grounded Q&A to the autonomous LoopAgent.
Real-time voice agents for voice support and answering calls, with three engines to match the scenario.
Orchestrate multi-step processes with 16 node types and call them via API, sync or async.
Each agent answers from its own knowledge base, with sources on every answer.
Choose the model for each agent and control its parameters, output format and cost.
Design agents visually — prompts, knowledge, tools and guardrails — and publish them without writing code.
Start with an Agent on your knowledge base. Use FlowAgent when a conversation must branch through set steps, Workflow when a process should run as an API, LoopAgent when open-ended questions need checking and tool calls, and AudioAgent to answer the phone.
Choose the model for each agent, with fallback models for Agent and FlowAgent. Tools connect via API, MCP or OAuth and are shared between GStudio and GWork; skills can be shared across the organization or attached to a single agent.
Publish to 25+ channels (website widget, shared page, API, DingTalk, WhatsApp, Slack, Teams, LINE, Messenger, Instagram and more) with 10 kinds of human handoff. LogTree traces every message across 57 node types, and version history plus Insights dashboards keep production under control.
Forward-deployed engineers build with your business team, then hand over a platform your team can operate.
Use data from GWork and Enterprise Search to identify recurring, high-value workflows your team understands well.
Choose the agent form and define knowledge, tools, guardrails and the human handoff path together.
Assemble in the studio, trace with LogTree, iterate on real cases, and publish once debugging passes.
Release to web, IM, API or phone channels with SSO, RBAC and PII anonymization in place.
Watch the Insights dashboards, restore a version if needed, and reuse the foundation for the next workflow.
Certified, auditable and available in your own data center.
Business workflows that became enterprise agents — co-built with forward-deployed engineers.
ManufacturingGlobalGStudio
ProfileA battery manufacturer selling through distributors and online channels in dozens of countries, with support requests arriving around the clock in many languages.
ChallengeSupport teams could not cover every time zone and language, and answers about specifications and warranties varied from agent to agent.
SolutionA customer-service agent built in GStudio on the product knowledge base, published to web and messaging channels with human handoff for complex cases.
ResultsMost routine questions are answered automatically, around the clock, at half the previous operating cost.
Public sectorHong KongGStudio
ProfileA department that reviews large volumes of submitted materials against published standards and guidelines.
ChallengeManual review was slow, depended on individual reviewers and made it hard to show why a decision was made.
SolutionAn AI review agent with the standards in its knowledge base and a Workflow that checks each submission, cites the relevant clause for every finding and routes exceptions to a human reviewer.
ResultsReview cycles shortened sharply, decisions became consistent, and every finding is traceable to a clause.
Customer serviceJapanGStudio
ProfileA large enterprise whose phone service center trains hundreds of agents a year on complex scripts and etiquette.
ChallengeRole-play training depended on senior staff, which limited how many agents could be trained and how often.
SolutionAn Audio agent plays the customer over the phone (inbound calls via Twilio) in realistic scenarios, while a Workflow scores each call against the script and gives feedback.
ResultsTraining runs whenever agents are available, with far less senior-staff time and a shorter path to the floor.
Start with AI for every employee. Connect AI to your company knowledge. Turn proven workflows into enterprise agents.
Five: Agent for knowledge-grounded Q&A, FlowAgent for conversations built from 13 components, Workflow for processes built from 16 node types and callable via API, LoopAgent that searches knowledge, queries data and calls tools until the request is resolved, and AudioAgent for voice support and phone calls.
