Archy AI Platform
Archy AI Platform: agents, knowledge and oversight for AI workflows
Build AI agent workflows, give them the tools and knowledge they need, and see what every run did, all on one platform. ArchyAgents is the workbench where you build them. ArchyPress, ArchySocial and ArchyAssistant are products built on the same platform, and you can build your own.
- ArchyPress
- ArchySocial
- ArchyAssistant
- MCP server
- OpenAPI
- Agents
- Workflow editor
- Tools via MCP
- Built-in RAG and knowledge management
- Memory
- Voice
- Observability
- Data sources
- Data pipelines
- Knowledge stores
- Advanced RAG
- Your RAG systems
- Identity
- Workspaces
- Model access
- Audit
Architecture
Five layers, from the workbench to the foundation.
The diagram shows how the parts of the platform relate. It is a concept, not a deployment drawing.
| Layer | Component | What it does |
|---|---|---|
| Workbench and products | ArchyAgents | The workbench and user interface for the Archy AI Platform: design, run and review agent workflows. |
| ArchyPress | Editorial publishing, built on the platform. | |
| ArchySocial | Social post studio, built on the platform. | |
| ArchyAssistant | AI assistant for Jira, built on the platform. Its dedicated agents run on the platform and are managed in ArchyAgents. | |
| Platform connectivity | Archy AI Platform MCP server | Lets any MCP-capable agent operate the platform. |
| OpenAPI | Connects other systems to the platform through a described API. | |
| Agent and workflow services | Agents | Each agent takes one job in a workflow and passes its result on. |
| Workflow editor | Lay out the steps, agents and handoffs of a workflow. | |
| Tools via MCP | Agents use software tools through the Model Context Protocol. | |
| Built-in RAG and knowledge management | Agents store information and retrieve it with simple retrieval-augmented generation (RAG). | |
| Agent memory | Short-term memory for the task in hand, and long-term memory that carries over. | |
| Voice interface | Work with agents by voice. | |
| Observability | Tracing, auditing and a history of agent workflow runs. | |
| Data and Knowledge (ArchyKnowledgebase) | Data sources | Files, databases, object storage and other systems, gathered through Dagster integrations. |
| Data pipelines | Orchestrated by Dagster: ingestion, transformation, and ontology and taxonomy analysis that map concepts and how they relate. | |
| Knowledge stores | A Neo4j graph knowledge base, Qdrant vector storage and conceptual snapshots. | |
| Advanced RAG | Gives Archy agents context built from concepts and relationships, not only similar text. | |
| Your RAG systems | Connect your own RAG systems or vector databases, one or several. | |
| Platform foundation | Identity | Sign-in and access for people who use the platform. |
| Workspaces | Separate spaces for teams on the platform. | |
| Model access | Access to the AI models that agents use. | |
| Audit | A record of actions on platform services. |
How it works
How a workflow runs on the platform
Five steps take a workflow from an idea to a run your team can inspect. Each step is carried out by a part of the platform shown in the architecture above.
Define
Set the goal, the steps, the agents and the handoffs before anything runs.
Connect
Give each agent or workflow the tools, systems and knowledge it needs.
Run
Agents carry out their steps and keep context in short- and long-term memory. People can work with them by voice.
Review
Edit the workflow, and start or stop it in ArchyAgents. Workflows can also be started through the API or the MCP server.
Observe
Follow each run through traces, audit records and run history, then improve the workflow in the editor.
Build on the platform
Your own AI solutions, on your data, with the models you choose.
ArchyPress, ArchySocial and ArchyAssistant show what can be built on the Archy AI Platform. Enterprises use the same platform to build solutions for their own processes, with their data kept separate and their own choice of models and knowledge sources.
Custom solutions, isolated data.
Build agents and workflows around the way your organisation works. Your data stays isolated from other customers.
One LLM provider or several.
Use the model providers that suit each task. The platform sits between your solutions and the providers, so a change of provider or model does not mean rebuilding what you built.
One RAG system or several.
Use the platform's built-in RAG and the advanced RAG of ArchyKnowledgebase, and connect your own RAG systems or vector databases through ArchyKnowledgebase — one or several, side by side.
Data pipelines to your agents.
ArchyKnowledgebase gathers knowledge from files, databases, object storage and other systems, maps how the concepts in it relate, and makes it available to your agents.
Agents that know more.
Expand what your agents know about your organisation, and their output becomes more accurate and more valuable.
Products built on the platform:ArchyPressArchySocialArchyAssistantCompare all products
How to evaluate
Three steps to a decision.
Bring one workflow.
In a technical walkthrough we look at the process, the tools and knowledge involved, and where people need to stay in control.
Try the products.
Use ArchyPress, ArchySocial, ArchyAssistant or ArchyStarter on your own work, or request access to ArchyAgents.
Agree the setup.
Decide whether the hosted products fit, or whether the platform should run in your own environment.
Enterprise
Run the platform in your own environment.
The platform and any of its products can be deployed as PaaS into your environment, with full data isolation and customisation for your processes. We scope each deployment with you.
Frequently asked questions
No. The Archy AI Platform is the set of services for designing and operating AI workflows. ArchyAgents is its workbench and user interface. ArchyPress, ArchySocial and ArchyAssistant are products built on the platform, and ArchyKnowledgebase brings data and knowledge from many sources to your agents. ArchyStarter is a standalone application in the Archy family that connects to ArchyAssistant.
ArchyAgents is the workbench and user interface for the Archy AI Platform. You design, run and review agent workflows there.
Through two interfaces. The Archy AI Platform MCP server lets any agent that speaks the Model Context Protocol, an open standard for connecting AI to software, operate the platform. Other systems connect through the platform's OpenAPI-described API.
Built-in RAG and knowledge management is part of the platform: agents store information and retrieve it with simple retrieval-augmented generation (RAG). ArchyKnowledgebase is a separate, more advanced Data and Knowledge Platform that integrates with the Archy AI Platform. It gathers knowledge from many systems, such as files, databases and object storage, maps the concepts in it and how they relate with ontology and taxonomy pipelines, and makes it available to Archy agents through advanced RAG pipelines, so agents work with concepts and relationships, not just similar text. It is built on Dagster, Qdrant, Redis and Neo4j. You can also connect your own RAG systems or vector databases to it, one or several.
Yes. Enterprises build their own agents and workflows on the Archy AI Platform, with isolated data, their choice of LLM providers and RAG systems, and data pipelines through ArchyKnowledgebase.
One provider or several. The platform sits between your solutions and the model providers, so you can change provider or model without rebuilding your solutions. We agree the providers for your setup in the technical walkthrough.
Yes. The platform and each product can be deployed as PaaS into your environment, with full data isolation and customisation. See enterprise deployment.
To explore an idea on your own, try ArchyStarter. If you already have a workflow in mind, request a technical walkthrough and we will look at it with you.
Dagster, Qdrant, Redis and Neo4j are trademarks of their respective owners. Their use does not imply endorsement.
Technical walkthrough
See how one of your workflows would run.
Bring a process, the tools it touches and the decisions people need to keep. We will show you how it maps onto the platform and what a first solution could look like.