Damper MCP Server

Post a timeline message and move the incident to investigating, identified, monitoring or resolved. resolvedAt is only valid with stage resolved. Private updates cannot change the stage of a public incident. isPublic and notifySubscribers default to true: public changes queue subscriber notifications and alert recipients are notified in the background. Success confirms the change, not delivery.

Post a timeline message and transition state (scheduled, in_progress, completed or cancelled). isPublic and notifySubscribers default to true. Private updates cannot change the state of public maintenance. Pass expectedState and expectedIsPublic from the latest read for conflict detection; reuse a requestId UUID for retries of the same post.

Add an alert recipient. type user_email and voice_call need userId (an account user; voice calls need a verified phone with urgent calls enabled), email needs email, slack_channel needs slackChannelId from the connected Slack workspace. Omit checkId to alert for all monitors.

Create an http, ssl or heartbeat monitor in a group. http and ssl monitors require a public url; heartbeat monitors return a heartbeatUrl for your job to ping. Probe header values are write-only and never returned (hasHeaders reports whether any are set). Creating an enabled monitor runs it once immediately. Plan limits apply to monitor count and interval.

Create a status page group (component). slug is lowercase letters, digits and dashes, unique per project. isPublic defaults to true.

Open a manual incident, optionally tied to a groupId or checkId. Dates use ISO 8601 UTC. initialUpdateMessage (or summary) becomes the first timeline update. isPublic and notifySubscribers default to true: public changes queue subscriber notifications and alert recipients are notified in the background. Success confirms the change, not delivery.

Create a scheduled status-page maintenance window. Dates use ISO 8601 UTC. Use list_status_maintenance_targets to choose groupId or checkIds (mutually exclusive). Omit both for all monitors when creating; omit both when editing to preserve scope. Set groupId and checkIds to null to target all monitors. isPublic defaults to true on creation. Public changes queue subscriber notifications for background delivery with retries. Success confirms the change and queue commit, not email delivery. Private windows still suppress monitoring alerts during the active schedule.

Permanently remove an alert recipient.

Permanently delete a monitor with its results and monitor-specific alert recipients. Set isEnabled false with update_status_check to pause instead.

Permanently delete a group together with its monitors and their results. Scheduled or in-progress maintenance for the group is cancelled.

Permanently delete an incident and its timeline. For a monitor incident, results recorded during it are cleared so uptime is recalculated. To keep history, post a resolved update instead.

Permanently delete a maintenance window and its timeline. To retain history, use add_status_maintenance_update with state cancelled instead.

Uptime history for a window (24h, 7d, 30d or 90d; default 90d) as bucketed points. Filter by groupId or checkId; without checkId only monitors included in uptime count. Maintenance is excluded.

Overall status page health: overall state, per-group state, monitor counts, active incidents and 24h/90d uptime (0-1, null without data). Maintenance windows show as maintenance.

List who is alerted when monitors go down or recover: account users by email or voice call, external emails and Slack channels, for all monitors or one checkId.

List monitors (status checks) with type, target, schedule, current state and last run. Set includeDailyUptime for 30 days of daily uptime per monitor.

List status page groups (components) in display order with visibility and monitor/incident counts.

List automated and manual incidents (open first, then newest) with affected component and timeline updates (newest first).

List maintenance windows, scope, schedules, visibility and timeline updates (newest first) for this project. Use the returned IDs to edit windows or timeline messages.

List status-page groups and monitors for maintenance targeting. Returns identifiers, names and visibility without probe configuration.

Run a monitor now and record the result. The outcome feeds the monitor state and may open or resolve an automated incident.

Edit an alert recipient, for example to pause it with isEnabled false or change its target. The same type requirements as create apply.

Edit a monitor. Passing headers replaces all stored headers; pass [] to clear them. http and ssl monitors require a public url; heartbeat monitors return a heartbeatUrl for your job to ping. Probe header values are write-only and never returned (hasHeaders reports whether any are set). Creating an enabled monitor runs it once immediately.

Edit a group name, slug, description, sort order or visibility.

Edit incident details, affected component, dates or visibility without notifying anyone. To change the stage, use add_status_incident_update.

Edit maintenance details, schedule, scope or visibility. For a state change with a timeline message, use add_status_maintenance_update. Use list_status_maintenance_targets to choose groupId or checkIds (mutually exclusive). Omit both for all monitors when creating; omit both when editing to preserve scope. Set groupId and checkIds to null to target all monitors. isPublic defaults to true on creation. Public changes queue subscriber notifications for background delivery with retries. Success confirms the change and queue commit, not email delivery. Private windows still suppress monitoring alerts during the active schedule.

Delete a context bundle and its version history.

Delete an executable routing fixture by name.

Delete a project-authored context relationship. Derived graph edges are not mutable through this tool.

Delete a structured context route by name.

Delete a project context section. Use when documentation is outdated or no longer relevant.

Get a context bundle with immutable version snapshots, optionally selecting one version.

Inspect the typed graph projection derived from context routes, sections, bundles, modules, and optional project-authored relationships.

Get full content of a specific context section. Supports glob patterns for hierarchical sections. **Hierarchical paths**: For monorepos or large projects, sections can be organized hierarchically (e.g., "api/architecture", "api/endpoints/auth"). Use glob patterns to fetch multiple: - `api/*` - direct children only (api/architecture, api/testing) - `api/**` - all descendants (api/architecture, api/endpoints/auth)

List versioned context bundles and recent versions.

List executable context routing expectations.

List the current deduplicated context-maintenance queue without refreshing or mutating it.

List project-authored typed relationships layered over the derived context graph.

List structured path-to-context routes in priority order.

List available project context sections with brief previews.

Upload/update a project context section. Use after analyzing local codebase. This action shares documentation with the Damper project for future AI agents. **When to use hierarchical names**: For monorepos or large projects, use paths like `api/architecture`, `api/testing` to organize by module. For simple projects, flat names like `overview`, `conventions` are sufficient. **criticalRules**: Short, actionable rules that agents MUST follow. These are surfaced automatically in `start_task` and `get_project_context` responses. ⚠️ SECURITY WARNING: Do NOT include sensitive information.

Add a property column to a page database. Property type. Relation types: `page` (linked page IDs), `roadmap` (linked roadmap task IDs), `feedback` (linked feedback IDs). Relation values are string arrays and sync PageLink backlinks.

Create a database-level page template with content and default property values.

Create a saved database view with table, list, board, or calendar layout.

Delete a saved database view. A database must keep at least one view.

Rename, hide, reorder, or change options/type for a database property. Property type. Relation types: `page` (linked page IDs), `roadmap` (linked roadmap task IDs), `feedback` (linked feedback IDs). Relation values are string arrays and sync PageLink backlinks.

Set database property values for a row page. Values are keyed by property ID. Date values are YYYY-MM-DD strings (for example, 2026-09-13); null clears a value.

Update a database view layout, visible columns, sort, filters, name, or position. Table and list layouts stack one list per value when config.groupByPropertyId is set; config.groupOrder is that group order. Board already groups into columns. Calendar does not stack groups.

Create a normal page, database page, or database row page. For rows, pass parentId as the database page ID and propertyValues by property ID.

Read a page, including live content, linked people, propertyValues keyed by property ID, and update timestamp for conflict-safe edits. Full data is included in text and structured output.

List pages that link to a page.

Read the complete database definition and every row, including property IDs, names, types, options and propertyValues keyed by property ID in text and structured output. Use only for small databases or an intentional full audit; prefer get_page_database_schema and query_page_database for growing databases.

Read all property IDs, names, types, options, views, and templates without downloading rows. Full definitions are included in text and structured output.

Read a bounded page of database rows using metadata or property filters, stable cursor pagination, a resumable polling cursor, and optional property/content projection. Custom date properties support comparison filters via propertyId and sorting via sort.propertyId (empty dates last). Dates use YYYY-MM-DD strings. Use updatedAt ascending (the default) for incremental polling; other sort cursors are for pagination only. Full projected rows are included in text and structured output. For array properties, in matches any supplied member while contains requires every supplied member. Prefer this over get_page_database for polling or growing databases.

Update page title, icon, content, content width, kind, agent visibility, or linked people. If ifUpdatedAt is omitted, the tool reads the latest page timestamp first.

Add a dependency between roadmap tasks. The blocker task must be done before the blocked task can be worked on. Provide either blockerId (this task is blocked by another) or blockedId (this task blocks another).

Create a new roadmap task. Use for importing from TODO files or creating work items. When the task comes from customer feedback, pass `feedbackIds` to link it in the same call: the feedback status follows the task and followers are notified, the same as linking later with `link_feedback_to_task`. Unknown feedback IDs fail the call before anything is created.

Create a roadmap task from a page. Copies the page title and content excerpt into the task, links the page as a source backlink, and updates a `roadmap` database property on the row when present.

Permanently delete a task. Only works on tasks in backlog or planned status with no commits.

Get task details. Defaults to a compact view (plan, current progress, subtasks, PRs, unresolved decisions, completion receipt, evidence counts) — no `agentNotes` dump. Pass `detailMode: "full"` for the legacy full payload (agentNotes, all sessions/commits inline), or request specific sections on demand with `include` (e.g. `["timeline"]` for composed history, `["checklist"]`, `["selfAssessment"]`, `["commits"]`, `["sessions"]`) while staying compact otherwise.

Remove a task dependency by dependency ID.

Update task fields like description, implementation plan, priority, effort, quarter, start/due dates, labels, or visibility. Pass status "closed" to discard a task that will not ship. Useful for refining task details as you learn more about the work. Set startDate and dueDate when planning so the task appears on the roadmap timeline.

Replace or append a stable context block while preserving unrelated blocks. The Markdown content must begin with a level-2 or level-3 heading whose generated slug exactly matches blockId.

Create or update a coherent context bundle. Semantic changes record immutable versions; identical retries are no-ops.

Create or update intended executable expectations for an important repository path. Derive assertions from project truth and verified code behavior, never by copying the current resolver output; only write project context when authorized, then run validate_context_routing.

Create or update an explicit typed relationship. Existing routes, bundles, modules, and section metadata remain authoritative; use relationships for missing cross-surface or semantic links.

Create or update a validated structured context route.

Create or update one database row by a parent-scoped canonical key. Retries with unchanged input are no-ops. Values are keyed by property ID. Date values are YYYY-MM-DD strings (for example, 2026-09-13); null clears a value.

Add completed tasks to a changelog. Private tasks can be linked; their titles/descriptions are not exposed in public changelog output and are surfaced as an "other updates" count.

Create a new changelog entry.

Permanently delete a changelog entry.

Update a changelog entry.

Get feedback details with votes, comments and a list of attachments. Download an attachment with `get_feedback_attachment`.

Download a screenshot or file attached to a feedback item. `index` is the 0-based position shown by `get_feedback` (`attachmentCount` in `list_feedback` / `get_feedback` tells you how many exist). Images come back as image content you can view; other files come back as an embedded resource with base64 data.

Get user feedback. Useful for understanding user needs. Feedback summaries include available dates such as `createdAt`, `updatedAt`, `firstOccurredAt`, and `lastOccurredAt`, and `attachmentCount` (screenshots or files the submitter attached; download them with `get_feedback_attachment`).

Update feedback status or visibility.

Delete a registered module by name.

Get details of a specific module.

Register or update a module in the registry.

Preview how a legacy flat completion checklist could be regrouped into native V2. Flags broad items, command-dump style items, and title-derived ids before writing the migrated document.

Parse Markdown routing into structured route candidates without writing them.

Preview a section replacement or stable-block patch, including semantic changes and sensitive-data warnings, without writing.

Get project context index (token-efficient). Returns section list with previews. Use get_context_section to fetch full content for sections you need. **IMPORTANT:** Call this at session start. Contains architecture, conventions, and patterns you MUST follow. Skipping this leads to mistakes and inconsistent code.

View current project settings, including the completion checklist index, standards refs, scope guidance, and the review evidence policy: per kind (apiExamples, screenshots, diagrams) a base level, the change kinds that trigger it, and rules that raise or lower it for specific surfaces, change kinds or risk domains; plus partialStageEvidence (partial hand-offs warn or block), plannedScope (whether prepare_work expected paths count) and checkDirectDone (whether moving straight to done is checked). Use `get_completion_checklist` to load detailed checklist branches.

Configure agent-relevant project settings, including the structured completion checklist and the review evidence policy (partial update: per kind a base level, optional trigger change kinds and `*Rules` that raise or lower the level for surfaces, change kinds or risk domains; plus partialStageEvidence, plannedScope and checkDirectDone — see the field descriptions). Preview the result with `get_completion_checklist` and a scope. Also sets `verificationCommands` (project-wide and per-surface commands prepare_work reports; Damper has no built-in defaults). This REPLACES the whole completion checklist document — use `patch_completion_checklist` to amend part of it. Prefer outcome-based proof groups, stable explicit ids, and referenced standards sections instead of command inventories.

List the API examples, screenshots and diagrams attached to a task, with the reviewer's verdict on each (approved / changes_requested + comment). Use it to check coverage before complete_task, and to find artifacts a reviewer asked you to redo.

Remove an API example, screenshot or diagram from a task, by id or by kind + key (e.g. a page that no longer changes).

Optional standalone self-review step. You can also pass `selfAssessment` directly to `complete_task` instead. Reflect on your session: what project knowledge was missing, what tools were confusing, what process gaps caused friction. This drives project improvements. If you have findings, propose improvements to the user after this call (e.g., `update_context_section` for missing docs, `report_issue` for tool problems).

Add a new subtask to a task. Useful for breaking down work into smaller steps. Optionally set startDate/dueDate (YYYY-MM-DD) so the subtask appears on the roadmap timeline.

Remove a subtask from a task. Use get_task to see subtask IDs.

Update a subtask: mark it done/not done and/or set its timeline dates. Use get_task to see subtask IDs. Dates are YYYY-MM-DD; null clears a date.

Bulk upload multiple modules at once.

Bulk upload all project context sections at once. Use after analyzing a codebase. ⚠️ SECURITY WARNING: Do NOT include sensitive information.

Bulk upload multiple templates at once.

Link feedback items to an existing roadmap task. Linked feedback status follows the task and followers are notified once the task is past backlog, the same as linking from the dashboard. To link while creating a task, pass `feedbackIds` to `create_task`. Already-linked and unknown IDs are reported back instead of failing the call; it fails only if none of the IDs exist.

Link the pull request that delivers this task. The PR is the primary deliverable of a task — call this as soon as the PR exists, and call it again with `state: "merged"` after merge. Upserts by URL, so repeated calls update the same PR record. **When to use:** - Right after opening a PR for the task (`gh pr create`): link it with url, branch, and title - After the PR is merged or closed: call again with the new state - Alternatively pass a `pr` object to complete_task if the PR was not linked earlier For GitHub URLs the repo and PR number are derived automatically.

List content pages and database pages for a project. Database row pages are included with their parentId.

Search pages and linked project knowledge by title/content.

Apply a scoped patch to the project completion checklist, touching only the branches you name. Groups, items, subitems, and standards are matched by `id`; entries with an unknown or omitted id are appended. Item ids are unique document-wide, so naming an item under a different group moves it rather than duplicating it. Listing groups/items/subitems only ever adds or merges — deleting is explicit, via the `remove*` fields. Anything the patch does not mention is left unchanged, so this is the safe way to amend taxonomy incrementally — use `update_project_settings` only when you intend to replace the entire document. Define `availableScopes` before writing context routes: route `appliesTo`, `changeKinds`, and `riskDomains` are validated against it, and unknown values are reported as health findings rather than rejected outright.

Apply stable-block operations with optimistic concurrency and immutable version snapshots.

Resolve final touched paths, compare them with the prepare_work snapshot, detect version drift and scope expansion, and generate context disposition evidence. Also returns `reviewEvidence` for the final scope: which screenshots, API examples and diagrams the project expects, how many are attached (since the last review decision, if any), and `missingRequired` gaps that would block `request_review` / `complete_task`. The scope is recorded on the task so the hand-off checks the same thing.

Build a deterministic, compact work packet with governing context, a shadow graph comparison and provenance, path-owned surfaces, checklist scope, project verification, unvalidated route-fixture representation, instruction drift, and guarded delegation guidance. `verificationCommands` come only from the project's own configuration (its verificationCommands setting, bundle verificationProfiles and quality-gate sections); when none apply the list is empty and `verificationGuidance` explains how to configure them. With `taskId`, also returns `reviewEvidence`: the screenshots, API examples and diagrams the project expects for this scope and how many are attached.

Report missing, irrelevant, or incorrectly routed knowledge without storing prompts or source code.

Report an error, bug, or feature request with MCP tools or suggest improvements. Reports that contain personal data, secrets, or chat transcripts are kept private automatically.

Resolve touched file paths to governing context sections, critical rules, risk domains, and suggested checklist scope. Uses routing/path-to-context when present and falls back to module paths plus section metadata.

Run bounded shadow-mode graph traversal for repository paths. Returns ranked context candidates and full provenance without changing authoritative route resolution.

Get the full content of specific blocks within a section. Use `get_section_blocks` first to see available block IDs, then fetch only the ones you need.

Get a block index for a context section. Returns headings with character counts. Use to selectively load only relevant parts with `get_section_block_content`.

Get full content of a specific code template.

Create or update a code template.

Release lock and return task to planned status. Works for in_progress and in_review tasks. **Before calling:** 1. Push any WIP commits **Summary parameter:** What was done, what remains, blockers. Helps the next agent.

Get the canonical Damper workflow instructions. Use this to set up or update CLAUDE.md in a project.

Attach a request/response example for an API endpoint you added or changed, so the reviewer can check the contract in Damper without running anything. Add one per changed endpoint (more for distinct cases, e.g. success and validation error — give each its own `key`). **Upserts by key:** the key defaults to "METHOD /path STATUS", so a success and each error case of one route are separate examples; calling again with the same key replaces the example. **Security:** use seeded/test data only. Authorization, cookie and API-key headers are always redacted. **Changed endpoints:** pass `beforeResponse` to show what the endpoint returned before your change.

Archive a page and its child pages. This is a soft delete.

Find invalid route taxonomy, weak or uncovered fixtures, missing baselines, route overlap, stale knowledge, and fallback usage.

List changelogs for the project. Filter by status (draft, published).

Log a commit for a task (secondary detail — the linked pull request is the primary deliverable, see `link_pr`). **When to use:** - After a notable commit: log hash and message - When completing work: optionally pass commits to complete_task instead **Example:** After `git commit`, call this with the commit hash and message.

Compare structured routing output with legacy Markdown routing for repository paths.

Mark task as ready for review (in_review). A human will verify and move it to done. Pass `skipReview: true` to mark as done immediately (use when the human explicitly asks). If the task is already in_review (put there by a previous complete_task call), calling this again will move it directly to done — no checklist required (already confirmed). Tasks put in review with `request_review` are the exception: wait for the reviewer, then resume with start_task. **Self-assessment (required):** Pass `selfAssessment` reflecting on your session — what project knowledge was missing, what tools were confusing, what process gaps caused friction. Pass empty object `{}` if nothing to report. For non-trivial code work, include at least one concrete entry instead of an all-empty assessment. Also pass `model` and `reasoningEffort` when known; top-level values override nested selfAssessment metadata. You can also call `review_task` separately before this, but passing `selfAssessment` here is simpler. **Before calling:** 1. Push all commits and open a pull request for the change — the PR is the primary deliverable 2. Link the PR with `link_pr`, or pass it here via the `pr` parameter 3. Check if project context docs need updating **Completion checklist:** If the project has a completion checklist (shown in `start_task` response), load relevant branches with `get_completion_checklist`, then resolve each active item by `itemId`. Use `status: "done"` with structured proof, or `status: "skipped"` for not-applicable items. If the checklist uses scope dimensions, pass `checklistScope` with explicit `appliesTo`, `changeKinds`, and `riskDomains` as needed. If a task crosses surfaces or contracts, bias toward broader scope rather than narrower scope. **Pull request:** Pass `pr` ({url, branch?, title?, state?}) if the PR is not yet linked. Commits are a secondary detail; the linked PR is what reviewers look at. **Delegations:** If you delegated work to sub-agents during this session, report them via `delegations` ([{agent, model?, description?}]). **Commits:** Pass commits array to log them at completion (convenience for final commits). **Completion receipt:** Pass `outcome` (defaults to `summary`), `deliveredScope` (defaults to titles of done subtasks), and `residualRisks` to build the compact completion receipt shown by `get_task`. This is what future agents and reviewers see instead of re-reading the full history. **Review instructions:** When sending to review (default), provide `reviewInstructions` — specific steps for the reviewer: what to test in UI, edge cases, areas of risk. Not a restatement of what was done. These are PUBLICLY VISIBLE on the roadmap while the item is in review, so write them for an external reader and keep credentials, internal paths, and private URLs out of them. **Review evidence:** When the change scope (declared `checklistScope`, scope inferred from `touchedPaths`, and what `prepare_completion` recorded) matches the project policy — `contract_changes` → API examples (`add_api_example`), `rendering_changes` / `public_output_changes` → screenshots (`add_screenshot`), and diagrams (`add_diagram`) when the project requires them — the evidence must be attached first. Missing suggested evidence comes back as `evidenceWarnings`; missing required evidence blocks the hand-off (error code `review_evidence_missing` listing each missing kind and the change kinds/surfaces that need it) unless you pass `evidenceNotApplicable` with the reason. After a review round, only artifacts attached since the reviewer's decision count. `prepare_completion` reports what is missing ahead of time. **Preview:** pass `previewUrl` when the change is running somewhere the reviewer can open (e.g. a Tailscale URL); internal only, never public. **Waiting:** when the task lands in in_review, the response tells you how to wait with `damper review wait <taskId>` (exit 10 = needs work, 0 = approved, 2 = timed out, 1 = error) if the Damper CLI is installed. Returns documentation update suggestions.

Load detailed completion checklist branches for the project. Use after `start_task` or `review_task` because those only return a checklist index. Filters: `groupIds`, `itemIds`, `appliesTo`, `changeKinds`, or `riskDomains` to load only the relevant subset. When scope is uncertain, bias broader rather than narrower. **Review evidence:** the response also includes `reviewEvidence`: the project's review evidence policy (base level, trigger change kinds and rules per kind) and, when `appliesTo` / `changeKinds` / `riskDomains` are given, the level each of API examples, screenshots and diagrams has for that scope and why. It is evaluated exactly as the hand-off gate (request_review / complete_task) does, so use it to preview what will be enforced.

Convert a candidate into a deduplicated private context-maintenance roadmap task.

Attach a Mermaid diagram that explains the logic or structure behind a change, so the reviewer can follow it without reading the diff. OPTIONAL — add one only when it clarifies something a list of endpoints or screenshots cannot: a decision tree or branching rule, a multi-step flow, a state machine, a call sequence, or how the touched modules, tables or classes relate. Skip it for simple changes. **Types** (first line of `source`): `flowchart TD` (decision trees, flows, dependency maps), `stateDiagram-v2`, `sequenceDiagram`, `classDiagram`, `erDiagram`. No `%%{init}` directives or `click` handlers. **Keep it small:** about 15 nodes or fewer, short labels, one idea per diagram. Quote labels that contain punctuation, e.g. `A["No tax line?"]`. **Mark what changed** in flowcharts and state diagrams by tagging nodes `:::added`, `:::changed` or `:::removed` (e.g. `B["Retry with backoff"]:::changed`) — Damper colours them; do not add your own `classDef` colours. **Before/after:** for a changed flow pass `beforeSource` with the previous diagram. **Upserts by key** (default: slug of the title); calling again with the same key replaces the diagram. Never put secrets or real customer data in labels.

Duplicate a page, including database row property values and linked people.

Correct a historical maintenance timeline message only. State and visibility remain immutable; this does not send subscriber notifications.

List registered modules in this monorepo.

Move a page under another page/database or reorder it. Use parentId null for a top-level page.

Add a note to a task, or resolve a previously reported blocker/open_question. Use ONLY for non-obvious approach decisions and blockers — do not restate commits, checklist confirmations, or self-assessment findings here; those are already stored structurally and shown in `get_task`. **Do NOT use for:** Session start/end ceremony, commit logging (use `add_commit`), status updates (use `update_task` with `currentProgress`), or restating what code changes do. **`kind` (default `observation`):** - `observation` / `decision` — appended to the free-text note log (Decision gets a **Decision:** prefix) - `blocker` / `open_question` — stored as a structured, resolvable event; the response includes `eventId`. Resolve it later by calling `add_note` again with `resolves: "<eventId>"` - `milestone` — stored as a structured timeline event, not resolvable **Templates** (copy and fill in): Decision: ``` **Decision:** Chose `libraryName` over `alternative` because it supports X and has fewer dependencies ``` Blocker: ``` Cannot proceed with `featureName` until: - Dependency X is merged - API endpoint Y is available ``` Observation: ``` Found that `componentName` already handles X via `methodName` — reusing instead of reimplementing ```

Get the public URL for a project page (roadmap, feedback, changelog, or a specific roadmap item). **IMPORTANT:** NEVER guess or construct public page URLs yourself. Always call this tool. **Pages:** roadmap, feedback, changelog, roadmap_item **roadmap_item** requires `itemId` parameter.

List all projects accessible with this API key. Use this to find project IDs for other tools. **Start here** to discover which projects you can work with.

Publish a changelog with optional email and Twitter notifications.

Aggregate privacy-safe feedback, self-assessment gaps, and fallback telemetry; may auto-create private tasks when project policy enables it.

Render the canonical structured route mapping as deterministic generated Markdown.

Reply to a feedback item as a team member. The reply will be visible to the customer. Send follow-up info to the feedback author by setting `notifyCustomer` (emails them) and optionally adding up to 3 `attachments` (PNG/JPEG/WebP/GIF screenshots, 3 MB each) that appear in the email and on the public feedback page.

Lightweight hand-off into review: move an in_progress task to in_review as soon as a human could review it (a working slice, an opened PR, attached screenshots or API examples, or when you are about to stop for the session). Do not leave reviewable work in in_progress. No checklist, PR, or self-assessment is required, and **no user confirmation is needed** — it is reversible and does not commit or merge. `complete_task` still needs confirmation. **Requirements:** the task must be in_progress and you must hold its lock (from start_task). **Effect:** status becomes in_review, the lock is released (so reviewers can act on it), and your session is closed. A reviewer approves or sends it back with feedback; resume with `start_task` if it comes back. **Stage:** `complete` (default) means the task is finished pending review. `partial` means a reviewable slice with more work to follow: approving it returns the task to in_progress instead of done. `complete_task` on a task waiting for a stage review requested here is rejected until the reviewer decides. **Review evidence:** the hand-off is checked against the project's review evidence policy for the change scope: `checklistScope` and `touchedPaths` passed here, unioned with what `prepare_completion` recorded; `prepare_work`'s expected paths count per the project's `plannedScope` (default: only when nothing else is known). Each kind's level comes from its base level and the project's rules; preview it with `get_completion_checklist` and your scope. Missing **required** API examples, screenshots or diagrams block it with error code `review_evidence_missing`, naming each missing kind and the change kinds/surfaces that need it. Attach them (`add_api_example`, `add_screenshot`, `add_diagram`) or, when they genuinely do not apply, pass `evidenceNotApplicable` with the reason (recorded on the timeline). A `partial` stage only warns unless the project sets `partialStageEvidence: "block"`; projects with `checkDirectDone` also check moving a task straight to done. After a review decision, only artifacts attached since that decision count, so each round carries evidence for what it changed. Approving the review re-checks the same evidence. Call `prepare_completion` first to see what is missing. **Review instructions:** optional steps for the reviewer. They are PUBLICLY VISIBLE on the roadmap while the item is in review — keep credentials, internal paths, and private URLs out of them. **Preview:** pass `previewUrl` when the change is running somewhere the reviewer can open (e.g. a Tailscale URL); it is internal only and never shown publicly. **Waiting:** if the Damper CLI is installed, run `damper review wait <taskId>` as a background command (exit 10 = needs work, 0 = approved, 2 = timed out, 1 = error).

Attach a screenshot of a page or screen you added or changed, so the reviewer sees the result in Damper. Capture every changed page in each relevant state (default, empty, loading, error, validation…), using the locale/theme/viewport the project asks for (see start_task). Give each state its own call. **Two ways to send the image:** - `image: { base64 }` — inline, fine for small images. - `requestUpload: true` — returns a one-time URL; then run `curl -X PUT -H "Content-Type: image/png" --data-binary @shot.png "<url>"`. Prefer this from a shell: it keeps large images out of your context. **Upserts by key:** the key defaults to page:state[:theme:locale:viewport]; reuse it to replace a capture. **Before/after:** for changed pages, send `beforeImage` (or `requestBeforeUpload`) to enable a comparison. **Size:** capture the viewport at 1–2× device scale, not the full page, unless the change sits below the fold. Damper re-encodes uploads to WebP (max 2560px wide) and makes a thumbnail, so bigger captures only cost upload time. **Privacy:** seeded/test data only — no real customer data. Max 5 MB per image, 60 screenshots per task.

List recent structured agent self-assessments for a project. Use this miss log to find recurring missingContext/toolIssues/processGaps and improve project context.

Lock and start a task. Fails if locked by another agent unless force=true. Use force=true to take over a task from another agent (e.g., if they abandoned it). Returns project context index with relevant sections for the task. If the reviewer decided your latest review request, the response starts with a `latestReviewRound` feedback block: address it before anything else. **IMPORTANT:** After starting, use get_context_section to read relevant sections. Project context contains architecture and patterns you MUST follow. **Workflow after starting:** 1. Read relevant context sections (architecture, conventions, etc.) 2. Do work on a feature branch 3. Open a pull request for the change and link it with link_pr — the PR is the deliverable 4. request_review as soon as something is reviewable (no user confirmation needed), or when stopping for the session 5. complete_task (done) or abandon_task (stopping)

Get roadmap tasks. Returns backlog/planned/in-progress by default. Search by title, description, plan, user story, or exact label, and filter by type, quarter, etc. Use `query` before paginating when checking whether a matching task already exists. Returns up to 50 tasks by default. Response includes `total` count. Tasks with incomplete blockers are excluded automatically. Task summaries include available dates such as `createdAt`, `updatedAt`, `startDate`, and `dueDate`. If total > returned tasks, paginate with offset (e.g. offset=50 for page 2). **Recommended:** Call `get_project_context` before starting work to understand codebase patterns.

List available code templates for this project.

Run project route fixtures against the production context resolver.

Record repository verification without mutating section content.