01/ 05
Customer support
Answers from your docs and past tickets, 24/7.
Chat · resolvedLive
- ✓Where is my order?
- ✓Arriving tomorrow before 6 pm.
- ✓Sources: Orders, Shipping policy
Example
Service 01
AI Agents
AI agents that answer customers, qualify leads and process documents, trained on your own data, with a clean handover to your team when a human is needed.
The stack
The tools behind AI Agents.
The platforms we connect to, and the developer tools we build with. We pick what fits your project and document all of it.
AI models & agents
OpenAIOpenAIClaudeClaudeGeminiGeminiGrokGrokDeepSeekDeepSeekMistral AIMistral AIMeta LlamaMeta LlamaPerplexityPerplexityOpenAIOpenAIClaudeClaudeGeminiGeminiGrokGrokDeepSeekDeepSeekMistral AIMistral AIMeta LlamaMeta LlamaPerplexityPerplexity
Developer tools
PythonPythonTypeScriptTypeScriptFastAPIFastAPICursorCursorVS CodeVS CodeJupyterJupyterPyTorchPyTorchDockerDockerPythonPythonTypeScriptTypeScriptFastAPIFastAPICursorCursorVS CodeVS CodeJupyterJupyterPyTorchPyTorchDockerDocker
Signature · inbox triage
Every message, sorted in seconds.
Watch the agent read incoming messages, decide what each one needs and how sure it is, then answer, draft or hand over.
Incoming · example6 waiting
Maya · Website chat
“Where's my order #2291?”
Intent: Order statusConfidence97%
→ routing to answered automatically
The agent reads every message, works out what it needs and how sure it is, then answers, drafts, or hands over with the full context.
Answered automatically0
Clear question, answer in the docs
Draft for review0
Needs a person to approve
Handed to a human0
Sensitive or unusual
Watch it work
One agent. Every conversation.
Pick a scenario and watch the agent handle it: it understands the request, uses your tools, follows your rules and replies in seconds.
Try a live agent in the Lab →Example · Website chat
1/4
How it works
From question to answer, live.
Every request follows the same path. Packets show the request moving through the system.
Now running
Answer
A clear reply in your brand voice, with the action already done.
Run log · exampledone ✓
- ●TriggerCustomer
- ●ContextAI agent
- ●AI stepKnowledge · Your tools · Guardrails
- ●ActionsAnswer · Your team
Under the hood
How an answer is made.
Each step is logged, so you can see exactly why the agent said what it said.
Example
My order arrived damaged. Can I get a refund?
Sorry about that! Order #48213 qualifies for a free replacement. I've started it, and the tracking link is on its way.
Total time
0.00s
- UnderstandIntent: refund request · confidence 0.97—
- Look upOrder #48213 · delivered Tuesday—
- KnowledgeReturns policy §2 · damaged on arrival—
- RulesRefund under limit · no approval needed—
- ActReplacement created · RF-90412—
- ReplyAnswer written in your brand voice—
- CheckNo personal data leaked · tone OK · sent—
What we build
Five agents, one platform.
Scroll through the jobs an agent can take off your team's plate.
02/ 05
Lead qualification
Asks the right questions and books the call.
New lead · score 86Live
- ✓Company: 3 clinics
- ✓Need: booking app
- ✓Booked: Thu 11:00
Example
03/ 05
Email & documents
Reads invoices and forms, fills your systems.
Invoice_Oct.pdfLive
- ✓Total: 1,240.00
- ✓Due: 30 Nov
- ✓Status: waiting for approval
Example
04/ 05
Voice agent
Takes calls, books and reschedules.
Call · 1m 12sLive
- ✓Rescheduled to Fri 15:30
- ✓Confirmation text sent
- ✓Summary saved to CRM
Example
05/ 05
Human handover
Hands tricky cases to your team with context.
Handed to SaraLive
- ✓Reason: complaint
- ✓Summary attached
- ✓Customer told: reply in 1 hour
Example
Industries
Built for how you work.
Connects to your tools
Works inside the tools you already use.
Inbox, chat, CRM, store and calendar. The agent reads and acts with the permissions you choose, nothing more.
Source connect
Docs, helpdesk, CRM and store, read with permission.
Action runner
Books, refunds, updates and replies inside your tools.
Approval gate
Anything risky waits for a person to say yes.
Audit log
Every step recorded, so you can see why it acted.
GmailWhatsAppHubSpotShopify
Google CalendarSlackNotionStripeZendeskOutlook
AI agent
Build · Launch · Improve
How we deliver your agent.
Trained on your business
- ✓We load your docs, products and past conversations
- ✓Set the rules: what it can do and when to hand over
- ✓Connect your tools with the permissions you choose
Knowledge sourcesExample
- Help-centre articles128 pages
- Product catalogue1,240 items
- Past support tickets6 months
- Refund rulesunder 200 auto
What you getHandover pack · AI Agents
- ✓Full source code in your accounts
- ✓Tests and documentation
- ✓Deployment to your hosting
- ✓Handover call and video walkthrough
- ✓30 days of fixes and tweaks
Yours from day one
Code · Docs · Accounts · Data
Built with
OpenAIClaudeGeminiGrokDeepSeekMistral AIMeta LlamaPerplexityHugging FaceLangChain
We choose tools that fit your project and your team, and document everything so anyone can maintain it later.
Studio work
Related concept
Our own product in this area, built the same way we'd build yours.
FAQ
AI Agents: common questions.
1Will it make things up?+
We ground answers in your own content, add guardrails, and route anything uncertain to a human.
2Can it use our tools?+
Yes. Agents can read and write to your CRM, helpdesk, email or database with the permissions you choose.
3Can we see one working?+
Yes. The chat in our Lab is a small example of a support agent.
Other services