description
No file selected
Machina
Online
Signed in as
No workflow starts (manual, schedule, webhook), running workflows stop before their next step, and paused ones stay paused. Chat stays available.
Enter an access token to continue.
Initializing...
Memory Usage
memoryLoading...
LLM Provider
smart_toyChecking...
Agent Fleet
groupsLoading...
Registered Tools
buildLoading...
LLM Usage & Cost · this session
paid$0.00
est. cost
0
total tokens
0
LLM calls
—
top model
No LLM calls yet — run a task with a provider configured.
Events Today
dynamic_feedLoading...
Pending Approvals
verified_userAll clear
Direct interface with Machina Core
Machina Intelligence
Neural link established. System status: NOMINAL. How can I assist you today?
LOCAL EXECUTION // INTENT-DRIVEN
CORE READY
All available system tools and their risk levels.
Browse memory scopes, preferences, and workspace profiles.
History of all executed plans and their outcomes.
Loading task history...
Create, edit, and manage reusable multi-step workflow chains.
Loading templates...
Loading workflows...
Cron & interval-based chain triggers. Runs automatically every 60 s.
Minimum 60 seconds
Loading schedules...
Manage specialised agents, capabilities, and inter-agent messages.
Pre-built agent profiles for common roles. Click to create a ready-to-use agent.
Loading agents...
No messages yet.
No threads yet.
Internal agent mechanics. These panels show how agents coordinate behind the scenes — voting, negotiating, and handing off work to each other. This happens automatically during task execution. The controls here are for inspection and testing.
No active delegation chains.
No consensus requests.
No negotiations.
No handoffs.
No agent requests.
No broadcast queries.
No collaboration sessions.
Agent learning metrics. These panels track how agents improve over time — capability scores, reputation, and consensus patterns are recorded automatically after each tool execution. Data here is read-only observability.
Select an agent to view learned capabilities.
Select an agent to view episodic memory.
Select an agent to view reputation.
No consensus patterns recorded yet.
Fleet Reputation Scores
Click to load performance comparison.
Result:
Agent observability. Health, memory, and capability data collected automatically as agents execute tools. Use these panels to inspect agent state — no user action is needed for normal operation.
Select an agent to view its memory.
No capability announcements.
No health data yet.
Select an agent to discover capabilities.
Click Run Discovery to scan all agents.
Select an agent to view health history.
Select an agent to view trends.
Click Refresh to load heatmap.
Workflow orchestration. Workflow chains and task queues are the production pipeline system. Chain execution history and the task queue are populated automatically when you run chains from the Studio or this panel.
No execution history.
No queued tasks.
Built-in Templates
Loading templates…
Steps (drag to reorder)
No workflow chains.
Select a chain to view its agent lanes.
handyman Tools
Messages
0
Negotiations
0
Pending Requests
0
Collaborations
0
Loading…
No communication events yet.
Select a negotiation to view its timeline.
Select a file to preview its contents
| PID unfold_more | Name unfold_more | Status | CPU % unfold_more | Memory unfold_more | Actions |
|---|---|---|---|---|---|
| Loading processes... | |||||
System Telemetry & Analytics
| Tool | Invocations | Success | Avg Latency | Min | Max |
|---|
System Event Stream
| Timestamp | Type | Source | Severity | Details |
|---|
Configuration & Preferences
—
Color Mode
Switch between dark and light theme
Accent Color
Density
Controls spacing and text size
Loading index status...
Secrets · Audit · Analysis · Scrubbing
Loading...
Loading audit log...
Model Context Protocol · Servers · Tools
No need to type tool names. Just describe what you want in the Neural Link:
External MCP clients (Claude Desktop, Cursor, …) can drive MachinaOS tools — risk-gated and audited. — tools exposed (LOW+MEDIUM).
MACHINA_MCP_ALLOW_HIGH is set). Unset it if this is not intentional.
stdio (local)
python -m core.mcp.serverSSE (remote)
GET /mcp/sseLoading servers...
Loading MCP tools...
Click Discover to load resources from connected MCP servers.
Click Discover to load prompts from connected MCP servers.
Site Visits · Sessions · Human vs Bot
Total visits
—
Human visits
—
Bot visits
—
Active sessions
—
Humans / 1h
—
Humans / 24h
—
Humans / 7d
—
Humans / 30d
—
Daily visits — last 30 days
Recent visits
| Time | Type | Session | User Agent |
|---|---|---|---|
| Enter admin key above and click Unlock | |||
Select an endpoint from the tree to explore its details.
Fetching from /openapi.json
Drop tools from the palette or double-click to add nodes
Connect output ports → input ports to build data flow
Navigation
Chat
Studio Canvas
Command Palette Modes
Preview
Type to search across all Machina resources…
Loading commands...
All tools will default to this directory
Tip: You can also type a path in the Browse tab's breadcrumb — it acts as an address bar.
Current: —
v0.1.0 — Local AI Operating Layer
Natural Language Commands
Type any request in Neural Link. Machina understands intents like "list files", "git status", "run the tests", or even "prepare my dev environment".
Safety First
Risky actions (delete, shell commands) require your approval. You'll see an approval prompt before anything destructive runs.
Full Audit Trail
Every action is logged. Check the Timeline view to see what happened, when, and why. Nothing runs in the background without a trace.
Dashboard
System status, quick actions
Neural Link
Chat interface, execute tasks
Tools
Browse all registered tools
Memory
Session, project, preferences
Workflows
Reusable multi-step chains
Agents
Multi-agent coordination
Timeline
Full task execution history with step details
> list files in the current directory
> git status
> show system info
> search for TODO in source files
Machina OS — Local-first, intent-driven operating layer
Press F1 anytime to open this guide
The intent-driven operating layer
Turn natural-language goals into safe, inspectable system actions. No scripts. No configuration. Just say what you need.
Browse repositories, read code, search for patterns, and analyze project structure — all from natural language.
Describe a goal and Machina decomposes it into a step-by-step plan, executes tools, and reports results.
Specialized agents collaborate — filesystem, git, shell, system — with negotiation, delegation, and consensus.
Every action is logged, every plan step is inspectable. Timeline, metrics, and events — nothing is hidden.
Session, project, and workspace memory. Machina remembers context across tasks and picks up where you left off.
Risk-aware policy engine. Dangerous actions require approval. Read-only operations run freely. Always reversible.
This is a sandbox demo. Explore, analyze code, and interact with AI-powered tools — safely.
Neural Link — natural language commands
Tools — browse available capabilities
Studio — visual workflow chains with conditional branching
Explorer — browse the demo project files
Write and delete operations are disabled in demo mode
An intent-driven operating layer that turns natural-language goals into safe, inspectable system actions.
Open Neural Link and type what you want to do
Machina creates a plan and executes it step by step
You approve risky actions — everything is logged in the Timeline