MotirBuilding in public
MOTIR · moooon
onMotir
You’re viewing a public project. Anyone can view it — no account needed. Sign in to submit, upvote, or comment on requests.View-only — you can’t edit work items
MOTIR-673

Epic 9: Native AI coding (hosted agent execution)

Not public
Description

The third layer of the pipeline made native — Motir runs the coding agent in a HOSTED cloud sandbox on the user's behalf. Until now Motir shipped only the external-agent (BYOK) form of AI coding (7.6 generates a prompt, the user runs it in their own agent — locally, optionally in the 7.9.7 sandbox container). This epic adds the hosted runtime the vision reserved as "a designed-for extension that augments — not replaces — the external-agent path" (MOTIR.md § What Motir is) — previously noted as a decided follow-up BEYOND the eight planned epics, now planned (2026-06-12, Yue: "plan one step further — host the container as the coding agent").

The hosted run mirrors the verified cloud-agent lifecycle (Devin / Google Jules / OpenAI Codex cloud / GitHub Copilot coding agent): a dispatched ticket → provision a container-per-run sandbox → the agent autonomously edits + tests → opens a PR → human review. Billing is usage-based, reusing the Epic-7 credit system: hosted CODING runs spend tokens and debit the SAME 7.12 credit ledger as planning (one balance covers both).

Builds entirely on Epic 7 — backward deps only: the 7.9.7 multi-agent sandbox image (the hosted image extends it), 7.6 dispatch (a hosted run is a dispatch-target variant), and 7.12 metering + credits (the token-usage report records an AgentRun and debits the ledger). Story 9.0 is the LLM metering gateway (fork of one-api) every hosted run meters through; Story 9.1 is the hosted-execution foundation — the hosted container + run-scoped auth (the user is logged in INSIDE the run) + gateway-metered token usage; Story 9.2 adds the runtime DESIGN-APPROVAL gate — distinct from MOTIR.md’s planning-time design gate: when the hosted agent PRODUCES a design in motir auto, a deployed preview (iframe) + a revise-chat hold the dependent subtasks on the user’s manual approval (per-project, default ON; the preview is undeployed on approval to cap cost). Future 9.x stories (deferred not forgotten): the PR review/iteration loop, multi-agent / parallel hosted runs, agent-selection policy, hosted-run pricing in the Epic-8 billing surface, and security hardening.

Child work items

This epic is not public

The project admin has kept this epic’s contents private.

Status
To Do
Type
Epic
Children
Hidden
Progress
Hidden