Chater

Your AI desk: native, private, and ready to do more than chat.

Chater is a fast desktop AI workspace for people who want their own models, their own data, and a practical path from a conversation to a completed task. It starts like a focused chat app, then grows with you into a tool-enabled, schedule-aware assistant—without trapping your work inside a browser tab.

Native desktop · Bring your own model · Local-first · Automate with confidence

Less browser clutter. More useful AI.

When an idea needs a quick answer, open Chater and start typing. When it turns into repeatable work, give the agent a tool, a workspace, or a schedule. Chater keeps that progression deliberate: you choose the model, choose what local files it may touch, and choose when an automated task may run.

If you care about…	Chater gives you…
Speed and focus	A lightweight native .NET and Avalonia desktop app with tray controls, global shortcuts, and no Electron-style browser shell.
Model freedom	OpenAI, Ollama, and OpenAI-compatible providers. Use a local model today and switch providers without changing your workflow.
Privacy you can inspect	Conversations, settings, attachments, and logs are stored locally in a user-controlled data directory.
AI that can act, not just answer	Optional tools for approved workspace files, the system browser, notifications, system information, scheduled jobs, and MCP connections.
Automation with boundaries	Scheduled agents run unattended with Docker or Podman shell access confined to one working directory.

From a thought to a finished workflow

1. Make every conversation feel like your own

• Stream responses with rich Markdown rendering, code blocks, links, and copied formatting.
• Send images to multimodal models from a file picker or directly from the clipboard.
• Create reusable skills for writing, research, analysis, translation, and the work you do repeatedly.
• Browse, reopen, and continue previous conversations instead of losing context in a pile of tabs.
• Choose system, light, or dark appearance and use English, Simplified Chinese, or Traditional Chinese.

2. Give the agent useful, explicit access

Chater keeps local access intentional. Select a file or folder to create a chat workspace, and the agent can read, write, create, and list only within that authorization boundary. Path traversal and symbolic-link escapes are checked before access.

When the task calls for it, the built-in tools can also:

• open an HTTP(S) page or an authorized local HTML file in the system browser;
• retrieve selected local details such as the current time, operating system, or signed-in user;
• manage recurring jobs directly from chat; and
• connect to configured MCP servers for additional capabilities.

3. Turn recurring work into a scheduled agent

Create a persistent job with a five- or six-field Cron expression, a prompt, and an absolute working directory. On schedule, Chater starts a fresh agent in unattended execution mode—ideal for recurring checks, reports, or project maintenance.

Scheduled agents are deliberately constrained:

• shell execution is Docker or Podman only, with the working directory as its boundary;
• direct workspace file and folder tools are not available to the scheduled agent;
• the agent does not stop to ask interactive questions; and
• concise milestones and errors are appended to chater-agent-job-<id>.log in the working directory.

4. Make sure important results get seen

Ask Chater to notify you now or on a schedule. Choose the delivery that fits the moment:

• Native uses the operating system’s notification system.
• TopmostWindow opens a user-dismissible Chater window that stays on top, adapts to its content, follows the active light or dark theme, and renders the notification body as Markdown.

Get started in a minute

1. Launch Chater.
2. Open Settings and add an API provider, or connect a local Ollama instance.
3. Select a model and a skill—or create a skill tailored to your work.
4. Start chatting. Attach an image, select a workspace when local files matter, or ask the agent to create a notification or schedule.

Your data, your controls

• API keys remain local and are sent only to the provider endpoint you configure.
• Conversations and messages live in a local SQLite database that you can inspect and back up.
• Application logs are written under logs/ in the data directory for straightforward troubleshooting.
• You can choose a custom data directory in General Settings and migrate existing local data to it.
• Tools are individually configurable, so you can keep an assistant conversational by default and enable action only when it is useful.

Run Chater locally

Prerequisites

• .NET SDK 10.0 or later
• A desktop environment supported by Avalonia
• Provider credentials or a locally running Ollama instance
• Docker or Podman only if you plan to run scheduled agent jobs with shell access

dotnet restore Chater.slnx
dotnet run --project Chater/Chater.csproj

On first launch, Chater creates its database and support directories in the operating system’s application-data location. Change that location at any time in General Settings.

Build, test, and publish

dotnet build Chater.slnx
dotnet test Chater.slnx --no-build

Release build:

dotnet build Chater.slnx --configuration Release

Current publishing targets:

• win-x64
• win-arm64
• osx-x64
• osx-arm64

See docs/release/build-and-artifacts.md for release and artifact guidance.

Project structure

Chater/
├── AI/            Agents, skills, tools, conversations, and model calls
├── Data/          SQLite access, repositories, and migrations
├── Jobs/          Persistent Cron scheduling and unattended agent execution
├── Services/      Settings, logging, localization, notifications, and platform services
├── ViewModels/    MVVM presentation logic
└── Views/         Avalonia windows and settings pages

License

See LICENSE for licensing information.