Desktop + Android companion · Public releases in preparation
Actual GlinkBot desktop app. A sample workspace, not customer conversations.
Features
More than a place to chat.
Models, tools, computers, and conversations. Connected in one workspace.
Choose a model per bot
Keep your preferred model for each bot, or change it for a single thread. Model and reasoning options follow the engine you connect.
Available models depend on your provider account and installed engine.
A computer when work needs it
Move beyond a text reply with supported browser, local VM, or cloud computer environments. Choose the environment in the bot’s computer panel.
Requires setup and a compatible engine. Local desktop control varies by platform; Linux Wayland is not supported.
Connect the tools you use
Configure connected apps or add trusted MCP servers to give compatible bots access to tools. Review each connection and its permissions before enabling it.
App connections need a configured connection service. MCP servers may run local commands or contact external services.
Keep conversations organized
Create threads and folders, pin important work, mark conversations unread, and archive what is finished. Give different projects their own space.
The screenshot shows actual bot and conversation controls in a sample workspace.
Configure your providers
Manage engine sign-ins, optional API keys, and OpenAI-compatible endpoints from your workspace settings. Choose the authentication method your engine supports.
Provider subscriptions, API charges, and usage limits are separate from GlinkBot.
Talk, don’t just type
Choose a voice engine, configure spoken replies, and decide when your bot should notify you. Keep voice and alert preferences together in its profile.
Voice calls currently require the macOS desktop app. Voice engines may need a separate account or local setup.
Files and contextBring relevant files into the task.
Web browsingResearch with a configured engine.
RoutinesGive repeat work a schedule.
A bot that feels yoursChoose its role, shape, and color.
Real app screens from a demonstration workspace. Feature availability depends on platform, engine, and configuration.
Use cases
Give the repeat work a home.
Start with a clear task. Add a schedule when it becomes a routine.
Example workflow
Morning briefing
Summarize what needs my attention today. Draft replies, but wait for my approval before sending.
Input
The messages, calendar, and task sources you choose to connect.
Action
Group updates by urgency and draft a short plan for the day.
Review
Check the priorities and every draft before sending or changing an event.
Schedule
Run on demand first. Add a weekday routine once the result is useful.
Example workflow
Invoice reminders
Review this invoice list. Flag overdue items and draft a polite reminder for each. Do not send anything.
Input
A spreadsheet or file containing invoice dates, amounts, and payment status.
Action
Prepare a clearly labeled reminder draft for each overdue item.
Review
Verify amounts, recipients, and payment records before contacting anyone.
Schedule
A weekly review, using the latest source file or an authorized connection.
Example workflow
Release checks
Compare this release with the acceptance checklist. Show what passed, what failed, and what you could not verify.
Input
Release notes, a test checklist, and an accessible test environment.
Action
Work through configured checks and collect evidence for each result.
Review
Inspect failures and unverified checks before approving a release.
Schedule
Run for each release candidate. Keep production changes separately approved.
Example workflow
Code review
Review this change for correctness and regressions. Give me concrete findings with file references. Do not edit the code.
Input
The repository and diff you authorize the coding agent to read.
Action
Trace the changed behavior and suggest targeted verification.
Review
Confirm each finding, then decide which changes to make.
Schedule
Start a new thread for each pull request or significant change.
Example workflow
Market research
Compare these products using their current official information. Link every source and separate facts from your interpretation.
Input
Your research question, comparison criteria, and allowed sources.
Action
Collect evidence and organize a concise side-by-side comparison.
Review
Check dates, sources, and gaps before using the conclusions.
Schedule
Repeat when you need an update; connected browsing tools are required.
Example workflow
Weekly reporting
Turn this week’s project updates into a short report. Highlight completed work, blockers, and decisions needed. Do not invent missing numbers.
Input
Project notes, task exports, and metrics you provide.
Action
Summarize progress, link supporting material, and flag missing data.
Review
Validate the figures and owner assignments before sharing.
Schedule
A weekly draft, with a human review before distribution.
Example workflow
Support sorting
Group these support messages by topic and urgency. Suggest replies and flag anything that needs a person. Leave all replies as drafts.
Input
An authorized support inbox connection or an exported message set.
Action
Categorize requests and prepare suggested responses.
Review
Check private details and tone. Handle sensitive cases yourself.
Schedule
A regular triage routine, limited to the sources you approve.
Example workflow
Personal planning
Help me plan a realistic week from these commitments. Protect focused work and leave some breathing room. Keep my calendar unchanged.
Input
Your goals, constraints, and an optional calendar connection.
Action
Suggest a manageable schedule and explain the tradeoffs.
Review
Choose what fits your week before adding or moving events.
Schedule
A weekly planning conversation, with a quick midweek check-in.
Examples depend on connected tools, permissions, and a running host. Always review important results.
How it works
Your machine is the starting point.
A desktop workspace. An Android companion. The AI provider you choose.
Your Android phone
Access your paired agents and conversations.
Paired connection
GlinkBot on your desktop
Keep the host running and reachable.
Provider requests
Your chosen AI provider
Your account, model, and usage limits.
01
Connect an engine
Configure a supported engine using its sign-in or API credentials. Start with the provider you already use.
02
Give each agent a role
Set its instructions, add useful context, and separate ongoing work into focused threads.
03
Keep control of access
Choose the tools and environment it needs. Review permissions, approvals, and the work it produces.
Workspace data uses ~/.glinkbot by default. Requests to cloud providers still leave your device.
Integrations
Bring the tools you already use.
Configure supported integrations in your workspace. Accounts and usage charges are separate.
Claude
Claude Code engine
Codex
OpenAI’s Codex agent engine
Cursor
Cursor agent engine
Grok
xAI provider integration
OpenCode
OpenCode agent engine
Kimi
Kimi agent engine
Qwen
Qwen agent engine
Droid
Factory’s coding agent
Hermes
Hermes agent engine
Pi
Pi agent engine
Antigravity
Antigravity agent engine
OpenAI-compatible
Your compatible API endpoint
Connect an installed engine or configure its supported provider access. Availability and model choices depend on your account and build.
Local VM
An isolated desktop on your machine
This computer
Local control on supported desktop sessions
Browser
A separately configured browser environment
Cloud / ASCIIBox
A configured remote computer environment
Self-hosted server
Run the workspace on a host you manage
Off
Keep computer access disabled
Not every engine supports every environment. Installation, credentials, platform support, and separate service charges may apply.
Gmail
Email context and actions
Google Calendar
Calendar context and events
Google Drive
Files and documents
Google Sheets
Structured spreadsheet data
Slack
Team conversations
GitHub
Repository and issue workflows
Notion
Pages and workspace knowledge
Linear
Issues and project work
Your MCP tools
Trusted local commands or server URLs
Examples from the connection catalog, not preconnected services or partnerships. Each needs authorization and a configured connection service; tool availability varies.
Compatibility does not imply partnership. Availability varies by platform and configuration.
Team playbooks
Start with roles. Build a team.
Turn a repeatable process into focused agents, shared instructions, and reviewable routines.
Project delivery
Go from a scoped idea to a reviewed change, with a clear handoff at each step.
No. It is a workspace for the agent engines and providers you choose. You configure the accounts, models, and tools that power your bots.
Does Android work on its own?
The Android app is a companion. It needs a paired, reachable desktop or server with the companion service running. It is not a replacement for that host.
Is every task offline?
No. Workspace data is stored locally by default, but cloud AI providers and connected tools receive the requests and context needed for the task. Their own terms and privacy practices apply.
Can I use my existing AI account?
Where the installed engine supports it, you can use its existing sign-in. Other engines use API credentials. Account eligibility, subscriptions, and usage limits remain with the provider.
Can an agent control my computer?
Only with a configured, supported environment and appropriate access. Browser, cloud, VM, and local desktop capabilities differ. Review permissions and approval settings before enabling tools.
When can I download it?
Public GlinkBot installers and the Google Play listing are being prepared. The download page will carry official release links when they are available. Pricing has not been announced.