Repository navigation
id(phase-14): Agents in worlds (MCP bridge) — v1 — first-iteration design + spike plan #6300
Description
Agents in worlds (MCP bridge) — v1 — first-iteration design + spike plan
Part of the Phase 14 core topic #6280 (Agents in worlds (MCP bridge)), filed from the next-horizons discussion #1240 per Drew's 2026-11-05 decision. The Phase 14 epic links the whole map.
Scope (rough outline — the core topic's research refines it)
MCP bridge spike: one agent joins a world as a client, reads screens, posts chat, performs one bounded program interaction — all under a capability ceiling with a full audit log. Deliverable: spike plan + design doc for agent identity/roles/guardrails + breakdown.
Entry condition
The core topic's research (#6280) has produced its first-pass findings, and the design-queue items ahead of it (see core issue) are moving.
Deliverable
A design document (in .opencode/plans/ or doc/) plus a breakdown of implementation issues for this step, linked from this issue and the core topic.
Drew's Phase 14 principles (#1240) govern this step: local-first always; hostable by anyone at any scale; p2p behavior identical to hosted behavior; our hosted instance is the public backstop, never a requirement.
Acceptance criteria
- Entry condition verified and recorded on this issue: the core topic research (id(phase-14): Agents in worlds (MCP bridge) — core topic (research + design) #6280) has produced first-pass findings, and the design-queue items ahead of this step are moving.
- A design document exists in
.opencode/plans/ordoc/covering: one agent joining a world as an MCP-bridge client, reading screens, posting chat, and performing one bounded program interaction — all under a capability ceiling with a full audit log — plus agent identity, roles, and guardrails, consistent with the Phase 14 principles from id: next horizons — what comes after Phase 13 (discussion with Drew) #1240 (local-first; hostable by anyone; p2p behavior identical to hosted; hosted instance is a backstop, never a requirement). - An implementation breakdown (linked issues) for the spike is filed and linked from this issue and the core topic id(phase-14): Agents in worlds (MCP bridge) — core topic (research + design) #6280, with native sub-issue/dependency relations where the platform supports them.
Activity
developing-today commentedon Nov 5, 2026
CODE-1865
dezren39 commentedon Nov 5, 2026
Intake review
🤖 Agent note · filed
Notes: This is the v1 (first-iteration) member of a four-part outline series (#6300–#6303) under core topic #6280. The later outlines (v2–v4) are deliberately thinner; this v1 carries the spike plan, so it is the one enriched in this pass. Nothing about the entry condition has been verified yet — the criteria make recording that verification the first step, so the design work cannot start silently before its gate.
Risks: Starting the spike before #6280's findings land re-litigates identity/guardrail questions the research is meant to settle.
Suggested approach: Verify the entry condition on #6280, then draft the design doc against the four principles, then file the breakdown as native sub-issues per protocol §11.
47 remaining items
dezren39 commentedon Nov 6, 2026
🤖 Agent note · handed off
dezren39 commentedon Nov 6, 2026
🤖 Agent note · handed off
dezren39 commentedon Nov 6, 2026
🤖 Agent note · handed off
opencode