ROADMAP

Where mcp-rune is going.
And what it isn’t yet.

This page is the project’s direction. Detailed planning lives in GitHub milestones and is tentative — items may be reshuffled or dropped without notice. Version numbers (0.2, 0.3, …) track production releases. The next upcoming version carries an estimated target date.

Legend shipped in progress planned researching blocked

↗ Milestones ↗ RFCs Become an adopter →

→ Next

UPCOMING

1.0.0 readiness

Exit criteria for committing @mcp-rune/mcp-rune to a stable 1.0.0 SemVer contract. Gated on a breaking-free soak of the post-#348 (0.107.0) API across all three consumers, plus resolving the explicit pre-1.0 surfaces and flipping to a post-1.0 deprecation policy. Tracked by the 'epic' issue in this milestone.

docs

1.0 launch: flip mcp-rune.dev messaging to "1.0 · stable"

Part of the #388 1.0.0 readiness epic. The pre-1.0 half is #392 (site work: mcp-rune/mcp-rune-site#57).

planned #393
docs

1.0 task: pre-1.0 stability messaging on mcp-rune.dev

Part of the #388 1.0.0 readiness epic. **Cross-repo:** implementation lives in mcp-rune/mcp-rune-site#57.

planned #392
—

1.0 blocker: reconcile stated vs. actual coverage thresholds

Part of the #388 1.0.0 readiness epic.

planned #391
docs

1.0 blocker: adopt a post-1.0 deprecation/versioning policy

Part of the #388 1.0.0 readiness epic.

planned #390
core

1.0 blocker: stabilize or @experimental-tag pre-1.0 public surfaces

Part of the #388 1.0.0 readiness epic.

planned #389

NOW SHIPPING

Pre-1.0 foundations

“Pre-1.0: extracting clean seams across tools, search, and apps”

Ten releases (v0.41 → v0.50) extracted the framework into composable layers. HttpExtension and ApiExtension frameworks split opt-in capabilities from core; custom-actions and search became standalone extensions; ApiClient and derived-fields moved to core; the DataLayer seam replaced BaseTool.apiClient; v0.50 closed the MCP apps overhaul with a convention-free seam, theming, and a bidirectional formatter registry.

apps

v0.50: MCP apps overhaul — close DataLayer leak, add styling/formatter extensibility, ship defaults

## Motivation

shipped #135
core

Introduce DataLayer seam; remove BaseTool.apiClient

## Motivation

shipped #129
extensions

Move static search to extensions['search'] bag (PR C, final step of search-extraction trilogy)

## Motivation

shipped #125
extensions

Move SearchService directory into the search extension (PR B of search-extension full extraction)

## Motivation

shipped #123
core

Relocate ApiClient and derived-fields to core (PR A of search-extension full extraction)

## Motivation

shipped #121
extensions

Extract search MCP tools as the second opt-in ApiExtension

## Motivation

shipped #119
extensions

Extract custom-actions as the first opt-in ApiExtension

## Motivation

shipped #117
extensions

Add ApiExtension framework: opt-in tools and ModelService mixins

## Motivation

shipped #115
extensions

Extract CIMD as the first opt-in HttpExtension

## Motivation

shipped #109

RESEARCHING

future

Ideas with no committed release. Some will become numbered items; others will be cut. Adopter votes shape priority.

prompts

Audit prompt strategies for v0.50 formatter patterns

## Motivation

researching

YOU

Shape it.

Open an RFC

For shaping items in this list. Discussed in the open, decided by the maintainers + founding adopters.

↗ github.com/mcp-rune/rfcs

Vote on milestones

Founding adopters get a binding +1/−1 per release. Three seats currently open.

↗ Read the adopter charter

File an issue

Bugs, papercuts, and small papercuts. Use the milestone field to suggest a release.

↗ github.com/mcp-rune/mcp-rune/issues

Where mcp-rune is going. And what it isn’t yet.

Where mcp-rune is going.
And what it isn’t yet.