Sees what you see Live
Understands screenshots, product mocks, documents, and visual artifacts, then explains findings in plain language you can act on immediately.
Built for practical AI execution
AI at your speed.Execution on your terms.
onDesk is your intelligent collaborator for real work. It can chat, analyze screenshots and documents, research the web, write code, and execute approved actions inside your project scope while keeping your context intact.
Model flexibility
Switch the brain, keep the flow.
Use the model that fits the task, from deep reasoning to rapid iteration, without changing your workspace.
Claude Opus 4.6Claude Opus 4.5Claude Sonnet 4.5
GPT 5.3 Codex
Kimi K2.5
GLM 5MiMo V2 Flash
Core operating model
What makes onDesk feel different.
These are the three behaviors that turn an assistant into a practical teammate.
Acts on your behalf Live
Runs commands, edits files, and executes workflows only with your approval, so momentum increases while control stays in your hands.
Understands your world Live
Keeps your project structure, recent decisions, and team patterns in context, reducing repetitive explanation and rework.
Detailed feature inventory
Capabilities across product, platform, and deployment.
A complete snapshot of what is live today and what is on the near-term roadmap, grouped by how teams adopt onDesk in production workflows.
Interactive feature matrix Live
Each card carries its capability tags so workspace, execution, MCP, enterprise, and roadmap context are visible without rigid category blocks.
Multimodal workspace canvas
Text, images, code, and web search stay in one continuous conversation flow.
Agentic execution controls
Toggle execution mode to run approved shell commands, file edits, and automation actions.
Project-scoped sessions
Attach chats to local repositories so the assistant understands directory structure and patterns.
Remote agent sessions
Create and recover remote sessions with explicit tool approval routing for controlled operations.
MCP server bridge
Connect custom tool servers so onDesk can interact with external systems over protocol contracts.
In-session model switching
Switch providers and models per task while preserving chat context and execution state.
Rich file ingestion
Processes PDF, DOCX, CSV, JSON, images, and code files as active context.
Context-aware web research
Injects relevant web findings into active tasks without leaving the workspace.
Mobile-responsive + remote-ready PWA
Expanded mobile-first access and remote workflows for on-the-go operations.
Deployment flexibility
Operate in cloud-hosted, self-hosted, or enterprise-managed deployment models.
Policy-aware admin controls
Govern model access, execution permissions, and operational boundaries per workspace.
SLA-backed enterprise channel
Dedicated engagement path for support, rollout planning, and operational assurance.
Deep Research engine
Systematic multi-source investigation with structured, citation-backed output.
Team workspaces + SSO
Shared collaboration space with enterprise identity integration.
Admin analytics suite
Usage analytics, spend visibility, and user management controls in one console.
Interactive functional artifacts
Proof in motion, not bullet points.
Three cinematic modules mirror how onDesk diagnoses, executes, and remembers context during real workflows.
Diagnostic Shuffler
Sees what you see: layered visual diagnostics auto-cycle through real troubleshooting scenes.
Telemetry Typewriter
Acts on your behalf: a live execution feed streams action-by-action with a terminal-style cursor.
Cursor Protocol Scheduler
Understands your world: cursor-guided day plan activates and saves context checkpoints.
Human judgment stays in the loop.
onDesk is built for high-trust, high-context work. It can reason, execute, and coordinate across tools, but critical intent still belongs to people who understand the outcome.
That means transparent actions, inspectable context, and predictable control flow whether you run solo, with a team, or inside enterprise governance.
"AI & Automation should raise standards, not hide decisions."
Layer 01 · Intake Mapping
Parse intent, classify dependencies, and flag approval gates before execution starts.
Layer 02 · Orchestration Loop
Coordinate model selection, memory retrieval, and tool invocation with strict response contracts.
Layer 03 · Verified Completion
Emit completion signals, preserve state snapshots, and expose audit-friendly outcomes.
Layer 04 · Operator Handoff
Package final results, pending actions, and recovery notes so the next operator can continue without context loss.
- OUT Summary + action log compiled Saved
- NEXT Pending approvals queued for review Ready
- SYNC Session context checkpoint archived Queued
Enterprise readiness
Deploy where policy requires.
Run onDesk as managed cloud or fully self-hosted infrastructure with access controls, model governance, and operational visibility.
- Self-hosted or managed deployment options across your own environment.
- SSO / SAML / OIDC support for centralized identity and role policy.
- Admin controls for model access, spend oversight, and execution permissions.
- Data sovereignty, auditability, and white-label-ready workspace branding.
Membership tiers
Simple structure. Serious capability.
Start with individual access, scale to teams, then move to enterprise controls when required.
Individual
Coming soon · Free tier
- Single user workspace
- All supported AI models
- Project-scoped chats
- Execution mode with approvals
- Community support
Team
Coming soon · Per seat pricing
- Everything in Individual
- Shared workspace + admin controls
- SSO authentication
- Usage analytics and budget visibility
- Priority support channels
Enterprise
Custom pricing
- Everything in Team
- Self-hosted deployment
- White-label and policy enforcement
- Data sovereignty controls
- Dedicated SLA and solution engineering
Early access contact
Contact us for early access.
We keep this simple: send your early access request to contact@onllm.dev. For faster review, include Use case, Name, and Contact number (optional).
Please make sure your message includes Use case so we can prioritize the right onboarding path.