Saved connections
Organise hosts, users, identity files and jump boxes in the remote-access manager.
Practical guidance for working across local shells, remote infrastructure, files and agent context from one workspace.
Symbio is currently distributed through the private beta on macOS, Windows and Linux.
Join the private beta from the main site and install the build supplied with your invitation.
Open a local folder. It becomes the shared home for your files, terminals and agent session.
Start locally or connect a saved SSH host, then decide whether that terminal should enter agent context.
Keep local shells, remote sessions and long-running processes visible in one grid.
Add panes for the shells and processes involved in the task, then arrange them around your workflow.
Choose the terminals that should receive a broadcast instead of sending every command everywhere.
Enable Context only on the panes the agent needs. Unselected shells remain outside the model request.
Save connection details once, move between hosts quickly and let the agent act on selected remote terminals.
Organise hosts, users, identity files and jump boxes in the remote-access manager.
Expose an SSH pane to the agent so it can inspect and execute commands on that machine.
See connection state and return to interrupted remote work without rebuilding the workspace.
Browse the connected file tree beside your terminals and agent session.
Use the same file workflow for local projects and connected SSH hosts.
Navigate the project tree and select a file from the local workspace or remote host.
Review and change code without leaving Symbio or losing the surrounding terminal context.
Attach only the relevant files or folders to the Neural Graph before asking the agent to reason about them.
The mode defines what the agent may do. Context defines what it can see.
Use OpenAI, Anthropic, Google, Groq, local models or an OpenAI-compatible provider.
Discover available agent and workspace actions with platform-appropriate shortcuts shown in the app.
Compose model context visually instead of sharing an entire workspace by default.
Create nodes from notes, terminals, files, folders, URLs, images or PDFs.
Group and link neurons so the graph captures how information belongs together.
Enable the neurons required for the task and inspect the final context before running the agent.
Connect external tools to the graph while keeping control over which operations the agent may use.
Choose an existing MCP connection or connect a supported account or local service.
Inspect read and write capabilities, then enable only the operations needed by the workflow.
Place the connection in a context group and enable it when the agent should use those tools.
Tools Studio is being built to turn natural-language instructions into reusable agent capabilities.
Explain its goal, expected inputs, ordered steps, constraints and successful output.
Review the generated instructions and refine how the agent should perform the task.
Choose when the new capability is available and which workspace context it may use.
Let the agent follow the same approved process whenever that tool is selected.