Runs where your team already works
Generic AI tools aren't built for your business
Off-the-shelf agents cost less to start and far more to live with. They don't know your data, they fragment across channels, and they hide what they're doing.
Generic AI doesn't know your data
Off-the-shelf models hallucinate your prices, miss your policies and have never read your knowledge base. Customers get wrong answers; your team cleans it up.
Every channel needs a separate setup
Website, Slack, WhatsApp, email. Each one is a new project, vendor and training cycle. Six months later you have five bots that don't agree.
No visibility into what AI is doing
Conversations vanish into a black box. No audit trail, no quality signal, no way to know if the bot is on-message, until something goes very wrong.
GuideLite solves all three.
From zero to deployed in three steps
Build a production-grade agent in an afternoon. No code, no data team, no new vendor cycle.
Build once, promote through governed environments
Every project starts with development, staging and production, and you can add your own up to five. Edits stay in draft, publish creates an immutable release candidate, and nothing reaches the next environment without gate evidence behind it.
Everything you need to build, ship and measure agents
One platform covers the full lifecycle. Stop wiring together five vendors.
A visual no-code agent studio
Drop in knowledge, choose a model, write your persona, set guardrails, ship. Versioning and rollback included.
Visual no-code builder
Drag, connect, configure. No engineers.
Every major model
OpenAI, Anthropic, Gemini, Llama, Mistral, Cohere.
Knowledge base & RAG
Document intelligence with smart chunking.
Role-based agents
Support, sales, HR, all on one platform.
Prompt studio
Versioning, testing, rollback.
A/B response testing
SoonCompare prompts in production.
The layer that gets you through procurement
Every prompt, model, tool and release is versioned. Every action is authorised. Every conversation lands in a tamper-evident audit chain. We publish our controls one by one, including the gaps.
Append-only audit trail
A per-org hash chain over every conversation with tamper detection. It fails closed in production rather than sign with a weak key.
ZNYX runtime guardrails
Input and output are evaluated against your policy on the ZNYX guardrails runtime, on the hot path, fail-closed in production, output buffered before release.
Traceability & observability
Every turn emits a structured trace of retrieval, prompt, model, tool call, guardrail decision, latency and cost, inspectable by trace ID.
FinOps for AI
An append-only cost ledger by agent, model, project and environment, so spend is attributable rather than a single monthly invoice.
Human approval gates
High-risk tool actions such as refunds, account changes and anything you flag pause for a named person to approve.
RBAC and scoping
Roles and org / project / environment scoping resolved on every service, not bolted on at the edge.
PII masking & redaction
Sensitive fields are detected and redacted in transcripts, logs and exports.
Versioned everything
Agents, models, prompts and tools are versioned, and each release pins the exact set that went live.
Compliance control mapping
Our controls are mapped to SOC 2, GDPR, HIPAA and ISO 42001 criteria, with each one marked implemented, partial or missing.
No third-party training
Your conversations are never used to train external models.
We hold no third-party attestation yet, and we say so plainly. Ask for our control list.
Every agent in the organisation, on one control plane
Not just the agents you build here. Register external agents too, then govern models, tools, prompts, knowledge, cost and releases from a single inventory, with traces and approvals behind every change.
Control-plane capabilities ship behind feature flags to design partners before general availability. The label on each card says where it actually is: Early access means built and test-covered but flag-gated.
Ten capabilities, one platform
Your agents come pre-loaded with the building blocks your team would otherwise stitch together.
Document intelligence
Ground answers in PDFs, contracts, manuals and spreadsheets.
Organizational search
Unified search across Notion, Drive, Confluence and Slack.
Image generator
Product images, banners and illustrations from text prompts.
Audio generator
Natural text-to-speech in 40+ languages.
Text summariser
Compress docs, transcripts, tickets and research.
Video to text
Auto-transcribe video and extract key moments.
Code assistant
SoonWrite, review and explain code inside the agent.
Data analyst agent
SoonQuery structured data in natural language.
Workflow automation
SoonMulti-step agentic execution across your tools.
Memory & personalisation
SoonPersistent user context across sessions.
See an agent answer, cite and escalate
A sample support agent grounded in Acme's knowledge base. Pick a question.
Know exactly what your agents are doing
Conversation analytics, CSAT scoring, cost per resolution and a full audit trail, in real time.
Built for every team that talks to people or data
One platform, role-specific agents. Start from a template or build from scratch.
Resolve tier-1 tickets automatically.
A support agent that knows your product, your policies and your tone, 24/7 on every channel.
Simple pricing. Scales with you.
Start free. Upgrade when you're ready.
- 1 project, 3 environments
- Up to 2 agents
- 1,000 credits included free
- Or bring your own model keys
- Web widget deployment
- Community support
- 5 projects, all 5 environments
- Up to 10 agents
- Own keys or credits, your choice
- All integrations and channels
- Analytics and trace inspection
- Priority support
- Unlimited projects and agents
- All 5 environments per project
- Your keys held in the credential broker
- RBAC, audit logs, PII redaction
- Human approval workflows
- Dedicated success manager