An open standard for defining AI context and collaboration across platforms

4 months ago 6

An open standard for defining AI context and collaboration across platforms

As AI models, agents, and humans increasingly collaborate across tools, platforms, and workflows, it becomes critical to define context, actors, data boundaries, and instructions in a way that is both machine- and human-readable. Disconnected systems and hidden rules often cause confusion, privacy risk, and inefficiency.

context.json solves this by providing a single, unified, and portable configuration file that precisely defines all relevant context, instructions, roles, and history for any AI-powered session, workflow, or collaboration. It makes context and collaboration explicit, composable, and auditable, enabling secure, robust, and transparent operations in any environment.

Open Standard: Freely available, community-driven, and vendor-neutral.
Cross-Platform: Works with any AI/ML model, agent, human, or service, across any stack or workflow.
Portable: Move, archive, or share entire sessions, instructions, and history as a single file.
Explicit Context: All instructions, sources, actors, privacy settings, and constraints are clear and inspectable.
Auditable Collaboration: Complete record of who did what, when, and why—critical for compliance, enterprise, and regulated use.
Extensible: Supports custom fields for domain- or use-case-specific configuration.

context.json Specification (v1)

The context.json standard defines a structured way to specify context, instructions, actors, sources, boundaries, and collaboration history for any AI-driven workflow. The goal: maximum clarity, security, and interoperability.

A unique identifier for the context (UUID or deterministic hash).

"context_id": "e3b0c442-98fc-1fc3-9fb3-3259aec0e27a"

The version of the context.json standard.

Short, clear statement describing what this context is for.

"purpose": "Real-time document co-editing with human and AI agents."

List of all human users, AI models, or services participating in this context.

"actors": [ { "actor_id": "[email protected]", "role": "user", "type": "human" }, { "actor_id": "gpt-4", "role": "assistant", "type": "AI", "details": { "provider": "OpenAI", "capabilities": ["summarization", "chat"] } } ]

External documents, APIs, or data used within the session.

"sources": [ { "type": "document", "uri": "https://example.com/project-spec.pdf", "description": "Project requirements" } ]

Explicit step-by-step instructions or rules for the workflow. Each is a clear, single string.

"instructions": [ "Summarize key requirements before generating code.", "Do not access external APIs without user approval.", "All output must be in markdown format." ]

Defines privacy, scope, output, and data usage constraints.

"boundaries": { "context_window": 8192, "privacy": "confidential", "output_format": "markdown", "other_constraints": [ "No personally identifiable information may be shared.", "Session expires after 60 minutes of inactivity." ] }

Chronological log of all significant events, messages, and actions.

"history": [ { "timestamp": "2025-06-21T12:00:00Z", "actor_id": "[email protected]", "event_type": "message", "content": "Started session." }, { "timestamp": "2025-06-21T12:02:00Z", "actor_id": "gpt-4", "event_type": "completion", "content": "Provided summary of project requirements." } ]

Creation or last update time for this context.

"timestamp": "2025-06-21T12:30:00Z"

Open-ended space for any additional configuration, protocol links, or metadata.

"extensions": { "related_protocols": [ "https://github.com/modelyaml/modelyaml", "https://llmstxt.org/" ], "custom_field": "example_value" }

{ "context_id": "d9f6c1b3-43e6-47c9-8eb8-ec0c7b6d900f", "version": "1.0.0", "purpose": "Cross-platform, human-AI collaborative code review.", "actors": [ { "actor_id": "[email protected]", "role": "reviewer", "type": "human" }, { "actor_id": "gpt-4", "role": "assistant", "type": "AI", "details": { "provider": "OpenAI", "capabilities": ["suggestion", "linting"] } } ], "sources": [ { "type": "document", "uri": "https://acme.com/repos/project/main.py", "description": "Python source file" } ], "instructions": [ "Review all code for security and efficiency.", "Suggest improvements as inline comments.", "Log every major suggestion to history." ], "boundaries": { "context_window": 4096, "privacy": "private", "output_format": "plaintext", "other_constraints": [ "No code changes allowed without user approval." ] }, "history": [ { "timestamp": "2025-06-21T10:00:00Z", "actor_id": "[email protected]", "event_type": "upload", "content": "Uploaded main.py" }, { "timestamp": "2025-06-21T10:01:00Z", "actor_id": "gpt-4", "event_type": "message", "content": "Suggested optimization for function process_data()." } ], "timestamp": "2025-06-21T10:05:00Z", "extensions": { "custom_label": "code_review_sprint42" } }

Eliminates hidden context. Everyone—human or AI—sees the same instructions, sources, and constraints.

Explicit boundaries and roles help ensure correct use of data, auditability, and regulatory compliance.

Move sessions and workflows between teams, tools, and AI agents with full fidelity.

A clear, inspectable record of all actions, sources, and events—enabling trusted, accountable AI.

Support for custom domains, fields, and future protocol links—built for growth and ecosystem integration.

context.json is an open standard, actively seeking community input and implementation feedback. See GitHub for drafts, specification, and examples. Discussion and improvements welcome.

context.json is an open standard ・ v1 ・ Contribute on GitHub

Read Entire Article