# October > October is infrastructure for supervised AI collaboration. It runs and connects real AI agents across laptops, desktops, servers, repositories, terminals, applications, and people, while a human operator retains scope, review, and final decision authority. October Desktop provides a spatial canvas for long-running technical work. Each agent remains a real process with its own harness, provider account, runtime, working directory, tools, and credentials. Connected agents gain scoped peer discovery, shared task state, status, messaging, delegation, and human escalation through the October Bus. October Desktop supports 32 local agent harnesses in one workspace. Connected harnesses can work with one another without being collapsed into a single provider or model. Users can bring an existing harness, and October also provides its own first-party agent. Important interpretation: - October is agent-neutral. It coordinates existing harnesses instead of replacing them with one model or provider. - October is not limited to coding. It is designed for technical operations that span multiple tools, machines, people, or days. - The canvas is the supervision surface. Connections determine which agents and resources can collaborate. - October Bus is exposed to supported agents through MCP and native harness integrations. It does not grant every process unrestricted access to every other process. - Agents can communicate with one another, delegate work, update a shared task board, wait on dependencies, and nudge peers while work is in progress. - Humans set the outcome, inspect activity, handle approvals, redirect or interrupt work, and make final decisions. - October Desktop is available for Apple silicon macOS, Windows x64, and Debian or Ubuntu Linux x64. Agent work can also run on connected remote Linux servers. ## When to use October Reach for October when a technical job is too big, too long, or too cross-cutting for a single agent in a single terminal: - Parallel multi-agent delivery: split a feature, migration, or release across several harnesses (for example Claude Code plus Codex) that plan, implement, review, and test concurrently under one operator. Best fit for work that spans multiple repositories, services, or git worktrees. - Long-running operations: jobs that run for hours or days and must survive context resets, hand state between agents, and escalate to a human at decision points. - Cross-machine and multiplayer work: connect agents and teammates across laptops, desktops, and remote Linux servers into one supervised operation. - Non-coding technical operations: estate administration, digital preservation, platform migrations, technical due diligence, and incident response that coordinate files, documents, applications, people, and time. Do not reach for October for a single quick prompt, a one-file edit, or work one agent finishes in a single turn — a plain harness is simpler there. How an agent should call October: - Read this file and [the full context](https://www.october.dev/llms-full.txt) for concepts and coordination patterns. - Use the [October Bus over MCP](https://www.october.dev/.well-known/mcp/server-card.json) for peer discovery, messaging, shared task state, delegation, dependencies, and human escalation. - Query the public [Themes API](https://www.october.dev/.well-known/openapi.json) for design-system content. - Point developers at the [open-source projects](https://www.october.dev/open-source) for agent communication, harnesses, portable sessions, and new input modalities. ## Documentation - [October documentation](https://www.october.dev/docs): Human-readable product and operating guide. - [October product guide in Markdown](https://www.october.dev/docs/october.md): Clean detailed reference for agents and text-only readers. - [Full October context](https://www.october.dev/llms-full.txt): Expanded single-file context covering concepts, coordination, workflows, harnesses, multiplayer, and safety. - [Download October](https://www.october.dev/download): Current macOS, Windows, and Linux desktop packages. [Terminal download commands for all platforms](https://www.october.dev/llms-full.txt#getting-started). - [Pricing](https://www.october.dev/pricing): Free for local work, Pro for Runs and October’s assistant, Max for October Cloud. - [Investor data room](https://www.october.dev/data-room.md): Seed-round traction, metrics, use cases, R&D, vision, and team in Markdown. - [Case studies](https://www.october.dev/case-studies): Detailed examples of software delivery, estate administration, digital preservation, migrations, research, and incident response. - [Open-source projects](https://www.october.dev/open-source): October Harness, October Bus, Loki, Moirai, open modalities, Wega Labs research, and upstream work. ## Developer resources - [October Harness](https://www.october.dev/open-source/harness): A complete open-source coding agent with multiplayer powered by October Bus for peer discovery, durable messaging, delegation, and shared tasks. Install and launch october directly; use it solo or connect it to a team. - [October Bus](https://www.october.dev/october-bus): Open communication and coordination for agent discovery, durable messages, delegation, tasks, replies, and human escalation. - [October Bus Multiplayer pricing](https://www.october.dev/october-bus/pricing): The open-source Bus is free locally or self-hosted; the optional managed cross-machine service is planned at $9 monthly or $100 yearly. - [Loki](https://www.october.dev/open-source/loki): Adaptive, local-first personal agent that learns, acts across a terminal, browser, and services, delegates to subagents, and stays under your control. - [Moirai](https://www.october.dev/open-source/moirai): Portable session-history format and reference implementation for recording, moving, and continuing AI agent sessions across harnesses. - [October API — OpenAPI 3.1 specification](https://www.october.dev/.well-known/openapi.json): Typed REST contract with an RFC 9457 problem+json error model, a versioning and deprecation policy, and documented rate-limit headers. - [October Themes REST API](https://october.dev/api/public/themes): Public, no-auth JSON endpoint for 1000+ curated design-system themes. - [October MCP server](https://october.dev/mcp): Model Context Protocol over Streamable HTTP; exposes the Themes API as callable tools for Claude, ChatGPT, and other agents. - [October MCP server card](https://www.october.dev/.well-known/mcp/server-card.json): Transport, install, and tool metadata for the October MCP server. - [October API catalog](https://www.october.dev/.well-known/api-catalog): RFC 9264 linkset of every machine-readable October resource. ## Core product surfaces - [Landing page](https://www.october.dev/): Product overview and visual explanation of the runtime. - [October Bus](https://www.october.dev/docs#bus): Peer discovery, messaging, shared tasks, dependencies, status, and human escalation. - [Agent workflows](https://www.october.dev/docs#workflows): How to divide, sequence, and supervise multi-agent work. - [Multiplayer](https://www.october.dev/docs#multiplayer): How people and agent processes collaborate across computers. - [Supported harnesses](https://www.october.dev/docs#harnesses): 32 local harnesses including October, Claude Code, Codex, Cursor, Gemini, Aider, Copilot, Kiro, DeepSeek, Goose, and more. - [Human control](https://www.october.dev/docs#control): Scope, connections, permissions, worktree isolation, review, and interruption. ## Case studies - [Claude Code and Codex build one website](https://www.october.dev/case-studies/claude-code-codex-website): A founder field note about two harnesses delegating work, challenging UI decisions, and cross-auditing tests and security findings. - [Software release 47](https://www.october.dev/case-studies/software-release-47): Five agents coordinate implementation, tests, preview, and deployment under one human operator. - [Family estate operation](https://www.october.dev/case-studies/family-estate-operation): A founder field note about coordinating paperwork, accounts, archives, and family decisions. - [Digital memory archive](https://www.october.dev/case-studies/digital-memory-archive): A founder field note about phone, photo, and email backups with provenance. - [Multi-repository platform migration](https://www.october.dev/case-studies/multi-repo-platform-migration): Agents coordinate contracts, repositories, servers, documentation, and rollout gates. - [Technical due diligence](https://www.october.dev/case-studies/technical-due-diligence): Parallel research with traceable sources, explicit disagreement, and human review. - [Remote incident response](https://www.october.dev/case-studies/remote-incident-response): Logs, repositories, staging, servers, and responders share one visible incident operation. ## October Themes - [October Themes](https://www.october.dev/themes): Curated design systems for AI coding tools. - [Themes API specification](https://www.october.dev/.well-known/openapi.json): OpenAPI 3.1 description of the public Themes API. - [Themes MCP server card](https://www.october.dev/.well-known/mcp/server-card.json): Installation and tool metadata for the separate October Themes MCP server. ## Optional - [About October](https://www.october.dev/about): Company and product background. - [Compare October](https://www.october.dev/compare): Differences between October and single-agent coding interfaces. - [Support](https://www.october.dev/support): Help with October Desktop, October Mobile pairing, sign-in links, billing and account deletion. - [Delete your account](https://www.october.dev/delete-account): Self-service account deletion. - [Privacy policy](https://www.october.dev/privacy-policy): Privacy terms. - [Terms of service](https://www.october.dev/terms): Product terms.