HydraticThe agentic AI for enterprises, ready to use.
Connect your existing applications and automate tasks using AI, with model choice, human approval, and deployment paths that match enterprise reality.
The approval gate,
streaming live.
Genuinely streamed, no backend calls. Nothing staged for this page.
Watch the agent run a real approval-gated task.
Everything needed to move AI
from demo to operations.
Connect your existing tools
Start with connectors across communication, knowledge, operations, developer, and observability systems. Custom connectors are available on request.
A memory that stays current
Hydratic remembers what matters — your preferences, projects, and context — and brings it into every conversation. Its bi-temporal memory engine knows the difference between what was true and what is true now, so when things change, the agent changes with you instead of repeating yesterday. Every memory stays visible, editable, and deletable.
An agent that shows its work
Give Hydratic a complex task and it lays out a plan before it acts — then works through it step by step, checking items off in real time. No black box, no waiting in the dark: you always see what's done, what's happening now, and what comes next.
Compatible with any model provider
Reuse existing subscriptions through BYOK, including Azure, AWS, Vertex, and other providers. Hydratic AI Gateway increases reliability and throughput when you need it.
AI Act friendly by design
Critical tasks can require human approval, and AI use stays transparent through clear run history, decisions, evidence, and audit trails.
Deep workflow integration
Trigger Hydratic agents from Slack, Teams, Discord, Google Chat, and WhatsApp so work starts where teams already operate.
Trigger agents from the places
work already begins.
1// apps/api: chat platform webhook2POST /webhooks/slack34// starts a durable workflow:5// the turn survives timeouts and reloads6startChatWorkflow(event)
Deployment choices for stricter
enterprise environments.
Hydratic can be consumed as SaaS, operated in your environment, self-managed, or paired with locally hosted open-weight models.
SaaS
Hydratic hosts and operates the application for you.
Hydratic Managed
Hydratic operates the application in your cloud or on-prem environment.
Self-managed
You stay in full control of operations with software licensed from Hydratic.
Local LLM hosting
Hydratic can host open-weight LLMs for organizations that need local control.
What's actually
shipped today.
GitHub actions available
Outlook actions available
of those actions require approval
chat platforms Hydratic runs in
Works with the tools you
already run, and keeps growing.
Hydratic reaches your team through chat platforms today, and acts inside a growing set of connected systems, with new connectors shipping regularly.
Trust is
non-negotiable.
Enterprise-grade governance is built into every layer of the platform, from how credentials are stored to how every action is reviewed.
Human approval for critical tasks
Every write action a connector can take is approval-gated by default, in the web chat and inside Slack or Discord.
Encrypted credential vault
Connector credentials are encrypted with AES-256-GCM and rotated automatically. Secrets are never echoed back.
Enterprise identity
Sign in with Microsoft Entra ID, with DNS-verified domain mapping so your organization owns its sign-in.
Transparent AI use
Every run keeps a clear history of decisions, evidence, and outcomes, all auditable end to end.
The approval gate, live in your workflow
Exactly what appears in the product. Nothing staged for this page.
The assistant wants to book flight HY101 for Ada Lovelace.
- Flight
- HY101
- Passenger
- Ada Lovelace
Working on it…
Request denied.
Every connector,
engineered the same way.
Our connector framework is internal: typed inputs, approval gates, and auth built in from the start. Need one that doesn't exist yet? We build it.
Hydratic developer resources: OpenAPI, public API, and llms.txt
TypeScript-native
Input schemas are zod, so validation and types come from the same source.
Writes gated by default
requiresApproval on any action pauses the run for a human decision.
Two auth shapes
staticCredentialsAuth for tokens, redirectAuth for OAuth. Pick what fits.
Lazy implementation loading
The manifest stays cheap; action code loads only when it actually runs.
export const outlookConnector = defineConnector({id: "outlook",name: "Microsoft Outlook",description: "Read and manage Outlook email and calendar events.",version,categories: ["communication", "productivity"],auth: [createMicrosoftGraphOAuthAuth({ scopes: OUTLOOK_SCOPES })],actions: [...mailActions, ...calendarActions],});
Research that shaped Hydratic
Hydratic's connector architecture follows published findings on tool selection, deferred loading, prompt injection, and curated examples — so agents stay accurate, inspectable, and approval-gated in production.
Tool-selection accuracy research
Tool-selection accuracy falls sharply beyond roughly 100 candidate tools. Pools under 30 stay above 90% accurate.
>100 tools
Anthropic: deferred tool loading, Opus 4.5
Deferring tool definitions instead of loading them all upfront raised Opus 4.5's tool-selection accuracy from 79.5% to 88.1%.
79.5% → 88.1%
Coding-agent security evaluation
Every coding agent tested was vulnerable to prompt injection, with adaptive attacks succeeding more than 85% of the time.
>85% injection success
Curated I/O examples vs. raw schemas
Curated input/output examples instead of raw schemas raised complex-parameter accuracy from 72% to 90%.
72% → 90%
Tool-selection accuracy falls sharply beyond roughly 100 candidate tools. Pools under 30 stay above 90% accurate.
Tool-selection accuracy research
>100 tools
Applied in Hydratic
SaaS plans for teams that
want to start quickly
Every SaaS tier includes out-of-the-box integrations. Choose the level of provider configurability, BYOK, and reliability your organization needs.
Pro
For teams starting with managed agent automation.
- All out-of-the-box integrations
- Predefined AI model provider
- Managed SaaS access
Business
For teams that need provider configuration and BYOK.
- All out-of-the-box integrations
- Configurable AI model provider
- BYOK available
Enterprise
For organizations with reliability, retention, and control needs.
- Fully configurable AI models
- Zero data retention options
- Hydratic AI Gateway reliability
- BYOK available
Every plan includes all out-of-the-box integrations. Token cost billed at provider rates.
Built for the future,
but the future is today.
Tell us where your agents should start: SaaS, managed in your environment, or self-managed.