Add memory to any AI agent with Vilix AI (MCP)
Add Vilix AI to an agent with compatible remote MCP support. Configure authentication, context retrieval, and saving before returning a reply.
An agent with compatible remote MCP support can connect to Vilix AI at https://api.vilix.ai/mcp. Authenticate and add the Vilix AI custom instructions or wire the read-and-save loop into your agent. The saved exchanges will appear in your Vilix AI account, where other tools connected to the account can access the relevant context.
How shared memory complements agent memory
A model request depends on the context it's given, but agent runtimes can maintain history, files, and persistent memory. OpenClaw and Hermes already do this, locally. Vilix AI adds a shared store so that relevant context from your other connected AI tools can reach the agent as well.
The Model Context Protocol provides agent runtimes with a standard way of calling external tools at inference time. The basic turn-by-turn pattern makes use of two Vilix AI tools: get_context and save_turn. The client must invoke them via instructions or orchestration; just registering the server isn't sufficient to establish the loop.
The three-step memory pattern
Configure a compatible agent to follow this pattern on each turn:
get_context (before acting): loads recent messages, saved memories, related past conversations, your user_rules, and current project and task state into the context window. The agent can use the retrieved history and instructions while working on the current request.
act: the agent does whatever it was asked to do. It uses the retrieved context to inform its tool calls, plans, and response.
save_turn (after composing the reply, but before sending it): persists the exact exchange to your Vilix AI account. The agent then sends the reply it saved and reuses the chat_id within the conversation.
This pattern is what enables cross-AI memory: the same store that a Cursor session writes to is the one that a Claude Code session reads from, and the one that your autonomous agent acts on.
Connecting the endpoint
The Vilix AI MCP server is available over Streamable HTTP at:
https://api.vilix.ai/mcpThe endpoint is the same for all clients, but they have differing configuration formats. Use your client's remote MCP settings and configure the authentication, then verify that a context read and a saved turn work. The essential connection information is:
Server URL: https://api.vilix.ai/mcp
Transport: Streamable HTTP
Authentication: OAuth in supported clients, or the supported API-key/Bearer setupAfter configuring the connection, authenticate, add the Vilix AI instructions, and then test retrieval and saving. A browser extension is not required.
For a step-by-step tour of the protocol and available tools, see the MCP integration docs.
When to use projects, tasks, and rules
Vilix AI memory has three surfaces that agents can read from and write to:
Cross-tool memory: a searchable store of past conversations and saved notes. A configured agent can retrieve relevant saved entries via
get_context. Useful for anything factual: decisions made, context established, work done.Projects and tasks: a light project manager that automatically injects its state into context. If your agent is managing a multi-step build, keep its project and task records up to date in Vilix AI so later runs can retrieve the saved state across connected tools and devices.
user_rules and project_rules: short directives that apply to every context load ("short answers", "always use TypeScript", "check open tasks before starting"). Context requests return available rules, which the agent must still follow.
These surfaces together mean that the agent doesn't just remember what happened. It also knows what is in progress and how you want things done.
Works across the tools you already use
Vilix AI works with agent harnesses and coding tools that support a compatible remote MCP connection and authentication. Examples include ChatGPT, Claude, Claude Code, Cursor, Codex, Grok, GitHub Copilot, Windsurf, Lovable, and purpose-built autonomous agent harnesses. See the following guides for deep dives on specific runtimes:
Each guide applies the same read-and-save pattern to a different agent workflow.
Getting started
The free plan covers your essentials. If you want the 7-day Pro trial with the memory features and project limits of the Pro plan, try Vilix AI free and connect your first agent in a few minutes. The MCP docs describe the protocol and link to supported setup paths.
Frequently asked questions
What is the Vilix AI MCP endpoint?
The endpoint is https://api.vilix.ai/mcp over Streamable HTTP. You add it as a custom MCP connector in your agent harness, authenticate once, and memory tools are immediately available. There is no SDK to install and no browser extension.
Does my agent need custom code to call get_context and save_turn?
Some clients can follow custom instructions without extra code; a custom agent might require explicit orchestration. Configure the endpoint and authentication, then instruct or wire the agent to: retrieve context, compose its reply, save the exact exchange, and return the saved reply. Verify that both calls succeed.
What gets loaded by get_context on each agent turn?
get_context returns recent messages from your history, semantically related past conversations, any memories you or a previous agent explicitly saved, your active user_rules, and the current state of any Vilix AI projects and tasks. Available context is retrieved before the agent composes its reply, with history and retrieval features depending on the plan.
Can multiple agents share the same Vilix AI memory?
Yes. Memory is associated with your Vilix AI account, not a specific tool or session. OpenClaw, Hermes, and Claude Code can connect to the same account. After one agent successfully saves an exchange, the others can retrieve relevant information from it.
Which agent harnesses does Vilix AI support?
A harness that supports a compatible remote MCP connection over Streamable HTTP. This includes ChatGPT, Claude, Claude Code, Cursor, Codex, Grok, Manus, GitHub Copilot, Windsurf, Lovable, OpenClaw, Hermes, and any custom agent runtime that supports the MCP spec.