moooon
Motir
Vibe the project. Bring an idea — Motir's three AI layers plan it, track it, and ship it, end to end. You're looking at Motir, built in Motir.
- Vibe Project
- Open Source
- AI Agent
- AI Loop
- 6
- requests
- 4
- upvotes
- 516
- planned
- 4,924
- shipped
- (motir-core) Every dispatched prompt tells the agent to call `report_action` before each step — the CLI's and hosted runs' prompt template, with the one-line rule and what an action may not containRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-skills) The `motir-run` skill's standalone procedure opens its run after the claim, calls `report_action` before every step, closes it on every exit path and names its harness and model honestly — parent runs included; the plugin ships the `mcp_tool` heartbeat hook; `SYNC.json` re-hashed against `run.md`Run provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-meta) `run.md` — a runbook run opens its run with `start_work_item_run` after the claim, calls `report_action` before every step, closes it with `close_work_item_run` on every exit path and names its harness and model honestly; a CLI-dispatched agent reports its steps too; and `install.sh` installs the Claude Code heartbeat hookRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core · test) E2E and acceptance video — a scripted MCP client claims a card, starts its run, reports its steps with report_action and a checkout, and closes it delivered; the run appears on `/runs`, in the run modal and in the card's run section with the agent, the model, its steps and a duration, and the card names its implementerRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core · test) Story integration gate — start → report_action steps and milestones → close over the real database for a leaf and a parent set, report_action inside a CLI run, the every-call heartbeat, the provenance stamp at a delivered close and none at a halt, the 60-minute reap ending at the last heartbeat, every refusal, and the coverage floorRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core) The MCP run tools — `start_work_item_run`, `report_action` (each step, milestones, heartbeat) and `close_work_item_run`, every Motir MCP call refreshing the caller's open runs, their permission keys and exemptions, the `DispatchRun` ruling rewritten, and `docs/mcp.md`Run provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core) `dispatchRunService` opens, reports into and closes an AGENT-REPORTED run — `reportedBy` on the run and on each event with its migration, the open over a claimed leaf or a claimed parent's children, `agent_action` steps resolved from the card key, the 60-minute agent liveness window, and the provenance a delivered close stampsRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- A card run from the `motir-run` skill is on the run record — the agent opens its run with its harness and model, reports its progress like `motir run`, keeps it alive, closes it on every exit, and the card records who implemented itRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- GithubAppTokenError: GitHub installation-token mint failed: token endpoint returned 503
- (motir-core · test) E2E — the built CLI against the real HTTP server with a fake agent that self-reports a model: `motir run` and `motir fix` legs read that model from `GET /api/v1/dispatch-runs/{id}`, and an agent that reports none leaves a null leg model on a run that still closesRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core · test) Story integration gate — over the real database: the back-fill migration on seeded legs (model, no model, malformed `data`, no event) and its second run a no-op, the append writing `model` onto the leg and leaving it on a later event without one, both reads returning it, and the coverage floorRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core) The CLI puts the agent's self-reported model on every leg it reports — `dispatchLeg` (next · run · run <scope> · batch · continue), `auto`, `fix` and `review` send `model` on `agent_exited`, null when the agent reported none, with an `@motir/cli` changesetRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core) Back-fill every leg written before the writer deployed — a forward-only migration setting `dispatch_run_card.model` from that leg's latest `agent_exited` event's `data.model`, under the writer's own validity rule, only where the column is still nullRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- (motir-core) `DispatchRunCard.model` and its writer — the column, the v1 append taking `model` on `agent_exited` (else the `data.model` an older CLI sends) and writing it onto the leg beside `exitCode`, the leg's model on the v1 run read and the browser's run DTO, and `dispatch-run-record.md` amendedRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- Every leg of a run records the model that ran it — `motir run`, `motir fix`, review and continue put the agent's self-reported model on the leg, not only inside an eventRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- Every message the person sends in the Motir AI window is labelled "turn 1", "turn 2", … — user bubbles show no label, just the message
- pagesMcpStoryGate case 4 needs both race orders in 24 runs but only a random 0–25 ms head start makes the markdown write win, so a loaded runner sees it lose all 24 and fails
- (motir-core · design) The decision port over a PAGE — the published version's text, number and author with Open page, the changed-since-published notice, the approved-and-frozen state, and the History panel's Frozen tag (delta mocks)Pages — a project's documents: write a page, keep it in the project tree beside the work it describes, and keep its history
- (motir-core · decision) An agent may report its OWN run over the MCP — amend `dispatch-run-record.md` and the `MCP_UNREACHABLE_RESOURCES.DispatchRun` ruling: the three run tools, `DispatchRun.reportedBy`, the events an agent may send, the provenance a delivered close stamps, and what stays the CLI'sRun provenance — every run of a card records who ran it, on which model and for how long, however it was started
- UnattributedMachineDestroyedError: Stopped Fly machine 48e1e4c3623248 (instance-2511b5b4-0558-4d45-bdff-70b5eab1987d) in motir-inst-e3ba1f50887009af, 11 min old: record_ended.