> ## Documentation Index > Fetch the complete documentation index at: https://docs.synti.co/llms.txt > Use this file to discover all available pages before exploring further. # Introduction > What Synti is, who it's for, and the core ideas behind the local-first desktop agent app. Synti is a local-first desktop agent that lives on your machine and does real work for you — writing and reviewing code, researching, drafting documents, and driving multi-step tasks across the tools and data you connect. Instead of a terminal prompt, Synti gives the agent a full desktop interface: an inbox of ongoing conversations, plans you can review before anything runs, and fine-grained control over what the agent is allowed to do. Everything runs on your own computer. Credentials are encrypted on disk in `~/.synti`, and Synti only reaches an external service when you explicitly connect it. There is also a headless server and a `synti-cli` binary for remote and programmatic use, so the same agent you drive from the desktop app can run on a machine you reach from anywhere. ## Who it's for Synti is built for people who want an agent embedded in real workflows rather than a chat window in isolation: * **Engineers** who want an agent that reads and edits a real repository, runs commands, and works with GitHub, Linear, and other services through connected sources. * **Operators and knowledge workers** who juggle many parallel tasks and want each one tracked, resumable, and documented. * **Teams** that need consistent, reusable behavior through skills, shared conventions, and permission rules that keep the agent inside guardrails. ## Core ideas A few concepts shape how Synti works. Understanding them makes everything else fall into place. ### Conversations as an inbox Every task is a [conversation](/core-concepts/conversations) — an independent thread you can start, pause, resume, and organize like email. Statuses, labels, and workspaces keep dozens of parallel tasks tidy, and each conversation's full history is preserved so you can pick up exactly where you left off. ### Permission modes You decide how much freedom the agent has. Three [permission modes](/core-concepts/permissions) — Explore, Ask to Edit, and Execute — range from strictly read-only to fully autonomous, and you can cycle between them at any time with **Shift+Tab**. ### Working directory and sources The [working directory](/core-concepts/working-directory) gives the agent direct, local filesystem access — reading, editing, searching, and running commands in a project folder. **Sources** extend the agent outward to external services and data, and come in three kinds: MCP servers, REST APIs, and local filesystems. ### Skills Skills are markdown instruction files (`SKILL.md`) that package reusable behavior — a review checklist, a build routine, a research procedure. The agent invokes them on demand, so complex tasks become a single mention. ## Bring your own model Synti is model-agnostic. Anthropic is the default provider, and you can also connect OpenAI, Google, OpenRouter, Groq, Mistral, xAI, or any OpenAI-compatible endpoint through a custom base URL. Connections are managed per workspace, and their credentials are stored encrypted alongside everything else. ## Next steps Download and set up the app on macOS, Windows, or Linux. Learn how sessions are created, organized, and resumed. Mentions, attachments, streaming, and steering the agent. Control exactly what the agent is allowed to do.