Skip to main content
Every task in Miteos runs inside a workspace — an isolated environment that gives agents a private browser, a file system, and a memory layer without interference from other work. Workspaces also let you group related tasks together so that agents can share context, files, and learned preferences across an entire project.

​
Workspaces

A workspace is a self-contained environment that persists beyond any single task. When you assign multiple tasks to the same workspace, agents can read files created by previous tasks, access shared memory, and build on prior results without you having to repeat context. Each workspace provides:
  • File storage — A persistent file system where agents read, write, and organize artifacts across all tasks in the workspace
  • Browser sessions — Isolated Chromium sessions scoped to the workspace, including cookies and authenticated states
  • Agent memory — Working, session, and persistent memory layers scoped to the workspace and its owner (see Memory System below)
  • Deliverables — All task outputs collected in one place, organized by the task that produced them
Common use cases for workspaces:
  • Organize all tasks for a single client or project under one roof
  • Share a data file across multiple analysis tasks without re-uploading
  • Maintain ongoing context for a long-running campaign so agents remember previous decisions
  • Keep personal automations separate from team workflows

​
Creating Workspaces

In the UI: Navigate to Settings → Workspaces → New Workspace. Give it a name and an optional description, then click Create. The workspace is immediately available when submitting new tasks. Via API: 
curl -X POST https://api.miteos.com/v1/workspaces \
  -H "Authorization: Bearer mt_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Q4 Marketing Campaign",
    "description": "All tasks for Q4 outreach, content, and reporting"
  }'
The response includes the workspace_id (e.g., ws_abc123). Pass this as workspace_id when creating tasks to run them inside the workspace.

​
Memory System

Miteos agents use a three-layer memory architecture. Each layer has a different scope and lifetime, letting agents carry just the right amount of context for the situation.

Working Memory

Scope: Single task executionHolds the active task’s context: current subtask state, intermediate results, in-flight tool outputs. Cleared automatically when the task completes. Agents use this to stay coherent within a task without polluting longer-lived memory.

Session Memory

Scope: All tasks within a workspacePersists across tasks as long as they run in the same workspace. Agents use this to remember decisions, data, and context from earlier tasks — for example, knowing which prospects were already contacted in a previous outreach run.

Persistent Memory

Scope: Per-user, across all workspacesLong-term storage of learned preferences, recurring patterns, and user-specific facts. Backed by a vector database with semantic search, so agents retrieve relevant memories by meaning rather than exact match. Grows over time as you run more tasks.

​
Managing Persistent Memory

Persistent memory accumulates automatically as you use Miteos. You can review and delete individual memories at any time. In the dashboard: Go to Settings → Memory to see a list of all your persistent memories. Each entry shows the content, the task that created it, and its creation date. Click the trash icon to delete any memory you no longer want agents to reference. Via API: 
# List all your persistent memories
GET /v1/memory

# Delete a specific memory by ID
DELETE /v1/memory/{id}
Deleting a persistent memory is permanent. If an agent relied on that memory for a recurring task, it will no longer have access to that context on the next run.

​
File Storage

Every workspace includes persistent file storage that survives across tasks and sessions. Agents can read files you upload, write new files during task execution, and organize outputs into folders. Key behaviors:
  • Agents read uploaded files — Drop a CSV, PDF, or image into the workspace and any agent in that workspace can access it as a tool input
  • Agents write files automatically — Generated code, reports, scraped data, and processed outputs are saved to the workspace file system without any extra configuration
  • Deliverables are files — Every task deliverable (dashboards, reports, exported data) is stored as a file in the workspace and is available for download from the Deliverables panel in the War Room
  • Files persist after task completion — Nothing is deleted when a task ends; you control the workspace lifecycle from Settings → Workspaces