Vector stores
Find relevant passages across your documents.
Documents → relevant evidence
Familiar APIs. Independent models. Persistent data.
The OpenAI Agents API.
Your choice of models.
Build agents through a familiar API, backed by managed vector stores, knowledge graphs, memory, and persistent sessions.
You build
Your application
Your interface, workflows, and business logic.
A familiar contract
OpenAI Agents API
Agents · sessions · turns · events
Choose the intelligence
Your models
OpenAI, Anthropic, and supported models through OpenRouter.
Explore models →Keep the context
Your agent data
- Vector stores
- Knowledge graphs
- Memory
- Persistent sessions
Connect the work
Your tools
MCP servers, hosted functions, and other agents.
Explore agent capabilities →Ragwalla manages the agent runtime and data layer behind your application.
01 / A familiar programming model
Start with the API
you know.
Create agents, open persistent sessions, and stream their activity through Ragwalla’s implementation of the OpenAI Agents API.
Read the API documentation →A familiar API. An Anthropic model.
POST /agents{
"name": "Research assistant",
"model": "claude-sonnet-4-6"
}POST /agents/sessions{
"agent_id": "<returned-agent-id>",
"environment": { "type": "none" },
"input": "Start a research session."
}Send these requests to your endpoint’s /gateway/v1 base URL with your Ragwalla API key. Use the first response’s agent ID in the second request.
02 / Model choice
Choose the model
for the job.
Configure the models your agents use and bring your own provider keys. Keep sessions, knowledge, and tool connections in Ragwalla as you evaluate which models work best for your application.
03 / A durable foundation
Give your agents something to build on.
Documents, relationships, remembered context, and conversation history. Each has a place in your application’s data layer.
Knowledge graphs
Connect entities and follow their relationships.
People · projects · relationships
Memory
Keep useful context across interactions, with explicit access controls.
Preferences · facts · shared context
Persistent sessions
Keep conversation history and the work performed during each turn.
Messages · turns · tool activity
04 / More capable agents
Put that foundation to work.
Tools & connections
Give agents access to your systems through MCP and hosted functions. Control which tools each agent can use.
Delegation & schedules
Let agents coordinate with specialists, run recurring tasks, and return to work at a scheduled time.
Applications & channels
Connect agents to your application, email, Slack, Teams, and other messaging channels.
Operate with control
Decide what agents
can access and do.
Set tool permissions and memory access. Track usage, configure retention, and inspect audit records as your application grows.
Build with Ragwalla