WebMCPGoogle
· Architecture · Roadmap · Learning path
Context Plane · Shared MCP/WebMCP review

One context engine. Multiple permitted ways in.

A person working with connected digital tools

A proposal to assess a reusable context engine across browser-native WebMCP, remote MCP, extension capture, files/connectors and manual input.

Status: PROPOSAL FOR REVIEW — NOT AN APPROVED DESIGN.
No adapter, tool contract, retention policy, deployment or integration path is approved by this document. Final decisions must be recorded in GitHub.

Recommendation for discussion

Keep one reusable context engine and expose it through adapters. The engine could normalise permitted inputs, classify and reconcile content, preserve provenance and freshness, identify trust and sensitivity, create task-aware context packs, validate them, and emit processing receipts. The first release should select the smallest combination that satisfies one measurable use case rather than implement every adapter.

Integration options

PathBest fitTrade-off / control
Browser-native WebMCPCompatible websites and clientsNot universally available; never make it the only path.
Remote MCP serviceReusable agent-facing tools across compatible clientsRequires hosting, identity, permissions and data-handling decisions.
Browser extension / user-mediated capturePermitted pages without WebMCPRequires explicit browser permissions and compatibility testing.
Files, clipboard and SaaS connectorsDocuments, exports and API-accessible systemsSource-specific authorisation and adapters.
Manual paste or uploadBroad baseline when automation is unavailable or inappropriateMore friction; source metadata may need user confirmation.

Suggested first fallback: manual paste/upload with source URL and capture time where available. Assess extension or connector support only after the first use case and permission model are agreed.

Proposed logical tool surface

context.search — search permitted sources for task-relevant context
context.capture — capture and normalise a permitted source
context.inspect — return provenance, evidence and classification
context.reconcile — link duplicates/variants and surface contradictions
context.pack.create — build a task-specific context pack within a budget
context.pack.validate — check permissions, provenance, freshness, conflicts and budget
context.pack.get — retrieve an existing pack
context.receipt.get — retrieve outcome and evidence for a processing run

These are proposed names, not existing endpoints. Owners must agree schemas, authentication, transport, versioning, errors and permissions before implementation.

Open decisions for MCP and WebMCP owners

  1. What is the first end-to-end use case and measurable success threshold?
  2. Should the first release expose remote MCP, browser-native WebMCP, or both?
  3. Which no-WebMCP fallback is in scope first?
  4. What is processed locally versus remotely, and what may be retained?
  5. How are users and agents identified, authorised, scoped and audited?
  6. What retention and deletion policy applies to raw captures, derived context, packs and receipts?
  7. Which fidelity, security and quality tests must pass before a context pack can be sent?

Acceptance tests

Working principle

Remove repetition, not meaning. Compress context, not evidence. Captured content is not automatically authorised instruction.

Related implementation proof: browser integration issue · durable queue and isolation issue. See also the architecture guide and release proof gates.