Agents and humans on the same chunks. How v1.5.0 made co-authoring possible.

Loomcycle v1.5.0 enables co-authoring between agents and humans by introducing user-scoped MCP dispatch, config-declared static principals, and deep-merge configuration. The update allows agents and humans to edit different chunks of a document concurrently under optimistic-revision concurrency, with Channel events firing on every chunk update.

loomcycle v1.5.0 ships three changes that compose into a workflow. RFC AG keys the /v1/ mcp dispatch off the authenticated principal, so user-scoped Memory, Documents, and Path under the MCP-server transport now share the same per-scope SQLite file as the off-run HTTP path the fix for the cross-transport mismatch where an MCP-created document was invisible in the Web UI . The route opens from substrate:admin to substrate:tenant; the per-tool gate inside the session still hides admin-only meta-tools. RFC AO adds config-declared static tenant, subject principals: a new principals: block in loomcycle.yaml binds a stable service identity to a token env name; the bearer secret lives in .env.local. One declared bearer authenticates both /ui/login AND an MCP thin client at the same tenant, subject by construction. RFC AN makes --config repeatable with deep-merge, so bundles stack onto operator config without copy-paste. What it unlocks: the launch plan moved from a flat Markdown file to a Document with 18 publication chunks; status is a queryable typed field; agents and humans edit different chunks concurrently under optimistic-revision concurrency; Channel events fire on every chunk update. The natural agentic shape: scaffold → drafter agent → human review → posted → reporter agent. Additive on the substrate layer; no new wire RPCs; TS + Python adapters unchanged at 1.4.0; Claude Code plugin bumps to v1.5.0. Existing deployments without a principals: block keep working unchanged.