Generative AI assistants perform best when they operate within well-defined parameters, but their effectiveness deteriorates when faced with unfamiliar codebases, proprietary security requirements, or institutional knowledge scattered across Slack conversations and internal documentation. GitHub Copilot Spaces addresses this limitation by enabling teams to curate specific context—source files, documentation, transcripts, and custom behavioral rules—into a single, reusable knowledge container. Once configured on github.com, Copilot interactions within that space draw from this curated information, producing responses that reflect organizational expertise rather than generic patterns. IDE integration for Spaces is planned for future releases.

The feature includes a five-minute setup guide for creating an initial space, options for customizing Copilot's tone and coding conventions through custom instructions, practical templates for accessibility, data queries, and employee onboarding, and a roadmap covering IDE integration and support for Issues and pull requests.

Why context is the new bottleneck for AI-assisted development

Large language models excel at pattern recognition but struggle with the irregular, context-dependent challenges that characterize real engineering work. These include monorepos combining modern frameworks with legacy code, organizational knowledge trapped in ephemeral communication channels, and security policies that diverge from public documentation. Without access to this context, AI assistants resort to assumptions. Copilot Spaces inverts this dynamic by allowing teams to designate which files, documents, and text snippets should inform Copilot's responses. As Kelly Henckel, PM for GitHub Spaces, explained, "Spaces make it easy to organize and share context, so Copilot acts like a subject matter expert." The practical outcome includes fewer incorrect suggestions, reduced manual copying and pasting, and code that requires minimal revision before committing.

What exactly is a Copilot Space?

A space functions as a secure, shareable repository combining knowledge artifacts with behavioral directives. It can contain code files, entire directories, Markdown documentation, transcripts, or plain text snippets that establish ground truth for Copilot's answers. Short system prompts within the space establish tone, coding style, or code review expectations. Spaces inherit the same role-based access controls already in use on GitHub. Attached files remain synchronized with the referenced branch, ensuring the space reflects current codebase state.

Spaces are accessible to anyone holding a Copilot license—whether Free, Individual, Business, or Enterprise tier—during the public preview phase. Administrators can activate the feature via Settings > Copilot > Preview features. Conceptually, a space operates like anchoring your team's collective knowledge to the Copilot sidebar and allowing everyone to query it using natural language.

Quick-start guide: How to build your first space in 5 minutes

  1. Navigate to github.com/copilot/spaces and select Create space.
  2. Assign a descriptive name, such as frontend-styleguide.
  3. Write a description clarifying when teammates should and should not use this space.
  4. Attach context by pulling folders (like src/components) or specific files (such as eslint.config.js) from repositories, or by pasting Slack threads, video transcripts, onboarding checklists, or JSON schemas into the Text tab, which Copilot treats identically to file attachments.
  5. Compose custom instructions—one or two sentences suffice—such as "Respond as a senior React reviewer. Enforce our ESLint rules and tailwind class naming conventions."
  6. Save and test by asking Copilot a question in the Space chat, for example "Refactor this component to match our accessibility checklist," and observe how it cites the attached files.

Personalize Copilot's coding style (and voice, too)

Custom instructions form the personality layer of a space, where their power becomes apparent because they coexist alongside content attachments. This combination enables significant customization through minimal text:

  • Enforce conventions: "Always prefer Vue 3 script setup syntax and Composition API for examples."
  • Adopt a team tone: "Answer concisely. Include a one-line summary before code blocks."
  • Teach project-specific vocabulary: "Call it 'scenario ID' (SCID), not test case ID."

During an interview on GitHub Checkout, Kelly described building a personal space for a nonprofit side project by attaching only the Vue front-end folder alongside instructions reflecting her preferred conventions. Copilot subsequently generated commit-ready code snippets matching her style guide on the initial attempt.

Automate your workflow: three real-world recipes

1. Accessibility compliance assistant

  • Markdown documentation covering WCAG criteria and internal "Definition of Done" standards
  • Custom instruction: "When answering, cite the doc section and provide a code diff if changes are required."

Rather than messaging the accessibility lead on Slack, teams can pose questions like "What steps are needed for MAS-C compliance on this new modal?" Copilot summarizes relevant checkpoints, references the documentation anchor, and suggests ARIA attributes or color-contrast corrections. GitHub's accessibility subject matter expert, Katherine, pinned this space in Slack so reviewers receive instant, self-service guidance.

2. Data-query helper for complex schemas

  • YAML schema files documenting 40+ event tables
  • Example KQL snippets stored as .sql files
  • Instruction: "Generate KQL only, no prose explanations unless asked."

Product managers and support engineers unfamiliar with database structures can ask questions like "Average PR review time last 7 days?" Copilot generates valid KQL queries with correct joins, enabling iteration without involving data science teams.

3. Onboarding Hub and knowledge base in one link

  • Key architecture diagrams exported as SVG text
  • Architecture Decision Records and design documentation from multiple repositories
  • Custom instruction: "Answer like a mentor during onboarding; link to deeper docs."

New employees can ask "How does our auth flow handle SAML?" and receive a structured answer with links and diagrams without leaving GitHub. Because spaces remain synchronized with the main branch, updates to architecture decision records propagate automatically, eliminating outdated wiki content.

Collaboration that feels native to GitHub

Spaces operate under the same permission framework already established on GitHub:

  • Personal spaces: restricted to the creator unless explicitly shared
  • Organization-owned spaces: controlled through repository or team permissions
  • Read-only vs. edit-capable: subject matter experts maintain authoritative versions while others consume the content

Sharing requires only sending the space URL or embedding it in a repository README. Anyone with access and a Copilot license can begin interacting immediately.

What's next for Copilot Spaces?

GitHub is expanding Copilot Spaces across additional workflows through several planned developments:

  • Issues and PR attachments to incorporate inline discussions and review notes into the shared context bundle
  • IDE Integration: Query Spaces in VS Code for activities like writing tests aligned with team patterns
  • Organization-wide discoverability enabling engineers to browse spaces as they would repositories, allowing new hires to search "Payments SME" and initiate conversations

User feedback will influence these priorities. Teams can share ideas and challenges through the public discussion forum or, for enterprise customers, via their account team.

Get started today

To begin using Copilot Spaces, visit github.com/copilot/spaces and follow these steps:

  1. Enable the preview feature: Settings > Copilot > Preview features > Enable Copilot Spaces
  2. Create an initial space with high impact—perhaps a code-review checklist or collection of common data queries
  3. Share the link via Slack or README and observe the reduction in direct messages to subject matter experts
  4. Refine iteratively by removing unused attachments, improving instructions, or subdividing large spaces

Copilot Spaces remains free during the public preview and does not consume Copilot seat allocations when using the base model. GitHub invites teams to explore what becomes possible when Copilot operates with precise, contextual information.

Source: GitHub Blog