AI agents built to work, not just answer.
Heralds gives companies the infrastructure to build, run and manage AI agents that understand their business and execute real work.
How Heralds works
From company context to real action.
From onboarding to active agents
Fast onboarding. Iterative expansion.
- 01
Establish company context
Onboarding starts by creating your initial company context in Core, giving agents the foundational knowledge to understand your business from day one.
- 02
Put your first agents to work
Start with ready-made workflows or define specific outcomes for your experts. Connect tools and channels (Slack, email, internal systems) as each agent needs them.
- 03
Iterate and expand quickly
Future work benefits from relevant knowledge and corrections, while agents take on more workflows over time. Start with one valuable workflow and expand from there.
Core
The right context. Without moving everything into another system.
Heralds Core maintains curated, shared context for your agents.
When a task needs additional information, Heralds can retrieve it directly from connected systems, without requiring a wholesale migration of your company's data.
Relevant context and feedback can inform future work, helping your agents operate with the knowledge and working practices they need.
Memory
Context that evolves with your work.
Heralds agents work with shared organisational context that can evolve over time.
Memories can be created on request or automatically when relevant context, playbooks or workflows need updating based on feedback.
This allows future work to benefit from relevant knowledge and corrections without requiring teams to repeatedly provide the same information.
Herald
One coordinator. Many specialised agents.
Herald coordinates work and helps create and configure Workspace Agents for specific tasks.
Each Workspace Agent can then operate independently, using the tools, permissions and shared context relevant to its work.
Workspace Agents
Specialised for the work. Independent in execution.
Workspace Agents handle specific tasks and workflows independently, with access to the tools, permissions and shared context they need.
Their AI models and reasoning effort are selected based on the complexity and requirements of the work.
They can run on demand, on a schedule or automatically when something happens.
Heralds Platform
The action layer connecting AI to your business.
Heralds brings together company context, agents, connected tools and workflows in one platform.
Agents can respond to signals, perform tasks and take action across your existing systems, with configurable permissions and approvals.
Your systems remain the source of business information. Heralds helps turn that information into completed work.
The Platform manages your workspace, people, permissions, connections, channels and agents, giving you one place to control how agents access information and take action.
Heralds Companion
Bring meetings into Heralds.
The optional Mac companion captures meetings without adding a bot to the call, turning conversations into structured notes and, when you choose, useful company knowledge in Core.
- No meeting bot
- Local privacy mode
- Multi-language
- Core integration
How it comes together
Built for agents that take action
- Manual, scheduled and event-based triggers
- Agent memory
- Connected tools
- Permissions
- Configurable approvals
- Actions
- Memories that update from feedback
How Heralds works
One platform. Four layers.
Heralds · System of Action
- Slack
- Web
- Companion
- Workspace
- Core
- Permissions
- Connections
- Agent configuration
- Conversations
- Workflows
- Tools
- Memory
- Actions
- Agent execution
Technology agnostic
The right intelligence for the job. Continuously evaluated.
Heralds selects AI models and reasoning effort based on the complexity and requirements of each workflow.
As new models become available, Heralds tests alternative configurations against existing tasks before adopting changes. This helps maintain output quality while identifying opportunities to improve speed and performance.
We keep up with AI, so you don't have to.
Turn your first use case into a working agent.
Start with one valuable workflow and expand from there.
