If you are switching to simse from another AI assistant, this guide helps you transition your workflow, move existing context, and map familiar concepts to simse equivalents. Examples use TUI slash commands — for headless mode, see the commands reference.
| Other assistants | simse equivalent |
|---|---|
| Conversation | Session |
| Chat history | Adaptive store + library |
| System prompt | System prompt builder + hooks |
| Custom instructions | Adaptive preferences |
| Plugins / extensions | MCP tools |
| API provider | ACP backend |
| Memory / recall | Evolving memory (adaptive store) |
The library builds up automatically as you work with simse. Context from your sessions is captured and indexed without any manual import step. You do not need to organize anything beforehand — the library system handles categorization automatically.
If you want to explicitly store something, ask simse during a session. For example:
Remember that our API uses JWT tokens with a 15-minute expiry.
simse uses the library_shelve tool to save the content with an appropriate topic. Over time, the library accumulates the knowledge that matters to your workflow.
To verify your context is being stored, ask simse to search the library:
Search the library for notes about deployment configuration.
simse uses the library_search tool to find matching entries and shows you the results. You can also ask it to list entries by topic using library_catalog.
Context carries over automatically. You do not need to paste previous context into new sessions. The adaptive store retrieves relevant entries based on what you are working on. When you start a session about a project, simse surfaces the most relevant entries from your history without you asking.
Backends are interchangeable. Switch between AI models through ACP without rewriting prompts or losing context. Your sessions and library are independent of the backend. If your current backend goes down, point simse at a different one and continue.
Tools are protocol-based. Instead of installing plugins from a marketplace, you connect MCP servers. Any MCP-compatible tool works with simse. Tool servers you already have (filesystem, database, API clients) plug in immediately.
Preferences are adaptive. Instead of writing custom instructions manually, simse learns your preferences from usage. You can still set explicit preferences, but the system adjusts on its own over time. The adaptive learning engine tracks which context is useful and adjusts relevance weights accordingly.
Sessions are forkable. When exploring a different approach, fork the session instead of starting over. /fork clones the conversation state into a new session. The fork starts from the same context and diverges from there.
The library grows automatically. You do not curate the library manually. As you work, entries accumulate. Near-duplicates are merged automatically. Entries are classified into topics. The adaptive store learns which entries are useful and surfaces them more often.
When you begin using simse, start your first session by telling it about your current projects, tools, and preferences. The adaptive store will use this as a foundation.
I'm working on a Rust backend for a document processing pipeline.
The codebase uses tokio for async, sqlx for database access, and
axum for the HTTP API. I prefer explicit error handling with
anyhow rather than unwrap().
This context is stored and carried into future sessions automatically.
Instead of keeping notes in separate documents, tell simse about them during a session:
Remember the key points from our architecture decision: we chose
event sourcing for the order service and CQRS for the read model.
simse stores this in the library using library_shelve and surfaces it in future sessions when relevant. You can also paste content directly into a session and ask simse to remember it.
Configure MCP servers once in mcp.json. They are available in every session:
{
"servers": [
{
"name": "filesystem",
"transport": "stdio",
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "~/projects"]
}
]
}Tools registered in mcp.json are available immediately without per-session configuration.