Everything applied AI needs, in one console.
AIS is where your agents are designed, your documents are understood, and your workloads run — in our cloud or on your hardware. One login, one API, one bill.
Multi-agent workflows, built in minutes.
Compose teams of agents that plan, call tools, execute code and hand work to each other — sequentially, in parallel, hierarchically or in loops. Build them visually, or let the AI Designer do it from a sentence.
AI Designer
Describe the outcome; the Designer drafts, builds, validates, test-runs and fixes the workflow — including its schedules, webhooks and knowledge bases.
Visual Workflow Studio
A full editor with undo/redo, agent duplication, import/export and Save & Run — every agent card open to inspection, nothing hidden.
Knowledge Bases
Index your documents into vector-searchable corpora agents query at runtime — grounded answers from your organization’s truth.
Agent Memory
Persistent memory scoped per agent and workflow, so your automations remember decisions, preferences and context across runs.
Schedules, Webhooks & Slack
Trigger from a cron schedule, an HTTP webhook, the API — or drop a photo and a message into a Slack channel and get files back in-thread.
Live streaming runs
Watch executions stream agent-by-agent over SSE — every tool call, artifact and hand-off visible as it happens, with full run history.
Documents in. Decisions out.
Define a document type once — its fields, types, formats and nested structure — then extract with field-level precision from anything your business receives. 18+ published document types come ready to use.
Model your schema
Nested fields with types, format instructions and expected values — built in a visual editor or raw JSON. Generic documents or fixed layouts, single files or bundles that need segmenting and classifying first.
Review what matters
Low-confidence extractions route to a review dashboard where AI-assisted corrections and structured editing make human sign-off take seconds, not minutes.
Deliver into your systems
Approved data fires your webhook with bearer, basic, API-key or HMAC auth — transformed to Finvoice, UBL or any XML/JSON shape your ERP expects.
Your workflows. Your hardware.
Everything AIS runs in the cloud also runs on devices you provision — GPU servers in your rack, industrial PCs on the factory floor, or a Raspberry Pi in the field.
One-line install
Linux (Ubuntu, Debian, CentOS, Fedora, Raspberry Pi OS — amd64 & arm64) and Windows 10/11/Server. The agent engine and local model runtime arrive as a managed stack that auto-updates with cloud builds.
Data stays local
Device installs can keep every document, extraction and review on-device — local storage, local models, local review — while templates and configuration sync down from the cloud.
Fleet visibility
Live health for every device — CPU, memory, GPU utilisation, VRAM, disk — plus a secure-relay log tail and per-device run history, right in the console.
GPU-aware, offline-capable
GPUs are auto-detected and used for local models. Policies govern how long a device may run offline before it warns and suspends — you stay in control.
$ sudo ./actgent install ✓ device provisioned · agent engine + local models running ✓ GPU detected — local models enabled → workflows, document types & knowledge synced from cloud
40+ tools. Every frontier model.
Agents come armed with built-in tools for search, code, documents and the apps your team lives in — and you choose the model behind every agent, per agent.
Search & Web
google_searchweb_searchtavily_searchread_webpageurl_contexthttp_requestCommunication
slack_send_messagesend_emailteams_send_channel_messageteams_send_chat_messageMicrosoft 365
calendar_create_eventonedrive_searchonedrive_create_fileonenote_create_pageDeveloper
code_executiongithub_create_issuenotion_create_pagejson_extractcalculateHuman & Control
request_human_inputexit_loopget_current_datetimeAnd beyond
Register your own HTTP endpoints as custom tools, plug in any MCP server, or call Aucterra platform services — Document AI included — as agent tools.
Aucterra MCP ServerGoogle Gemini
platform-managed — works out of the box- Gemini 3.5 / 2.5 / 2.0
- Imagen 4 (images)
- Veo 3.1 (video)
- Native audio & TTS
OpenAI
bring your own key- GPT-5
- GPT-5 Mini
- GPT-4.1
- GPT-4o
- o3
Anthropic Claude
bring your own key- Opus 4.7
- Sonnet 4.6
- Haiku 4.5
Local — on your device
local runtime, no API key- Llama 3.1 / 3.2
- Qwen 2.5
- Phi-3 Mini
- Gemma 2
- any open-weight model
See it with your own workload.
Start free and run a template on your own documents — or book a walkthrough with an engineer who built the platform.