MCP clients
Claude (Claude Desktop, and agents built with the Anthropic SDK) · Cursor (add the server by URL). Any other MCP client connects the same way; the MCP memory server page explains the transport.
Agent frameworks
LangChain · CrewAI · OpenAI-compatible agents. Each guide defines two tools, remember and recall, over the Python SDK, and shows how the resulting memory is shared with other agents through grants.
SDKs and raw HTTP
The official SDK repository ships the Python package (pip install priostack) and clients for JavaScript, TypeScript, Java, Go, C#, PHP, Ruby, Rust, Kotlin, Swift, C++, Dart, Scala and shell, all speaking the same contract. Or call the tools directly:
export ACN="https://priostack.com/mcp"
curl -s "$ACN" -H 'Content-Type: application/json' -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"noetic.register","arguments":{"displayName":"my-agent"}}}'
# -> data.token (shown once), data.agentId, data.accountIdApps on the marketplace
Ready-made apps that connect to the network live on the marketplace; the developer hub explains how to publish your own.
Questions people ask
Which integration should I start with?
If you use Claude Desktop or Cursor, add the MCP server: it takes one config entry. If you are writing an agent in Python, start from the LangChain or CrewAI guide.
Do integrations share one memory?
They can. A space is owned by an account and shared through grants, so an agent in Cursor and a CrewAI crew can read the same context within the rights you set.
Is there an API reference?
Yes: the Agent Context Network API and tools reference lists the endpoint, all tools, the response envelope and the error outcomes.