EN HR SL
Start Free

Damper MCP reference

Damper exposes 100+ tools over a hosted MCP server with streamable HTTP transport — no local installation. Agents use them to work tasks, resolve context, read pages and databases, act on feedback and publish releases.

Setup

Add the Damper MCP server to your AI assistant. It uses streamable HTTP transport — nothing to install locally.

Claude Code

Add to your project's .mcp.json or global ~/.claude.json:

// .mcp.json or ~/.claude.json
{
  "mcpServers": {
    "damper": {
      "type": "http",
      "url": "https://api.usedamper.com/mcp",
      "headers": {
        "Authorization": "Bearer dmp_..."
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root:

// .cursor/mcp.json
{
  "mcpServers": {
    "damper": {
      "type": "http",
      "url": "https://api.usedamper.com/mcp",
      "headers": {
        "Authorization": "Bearer dmp_..."
      }
    }
  }
}

Authentication

All requests are authenticated with your API key in the Authorization header. Find your key in the dashboard under Settings → API Keys — keys use the dmp_ prefix.

Authorization: Bearer dmp_your_key_here

Recommended workflow

  1. Prepare the work

    get_agent_instructions sets up your CLAUDE.md; prepare_work builds a compact work packet with governing context and graph comparison.

  2. Load context

    get_project_context or resolve_context — architecture, rules and the sections that govern the files you're about to touch.

  3. Pick and lock a task

    list_tasks sorts by weighted demand; start_task locks it and returns the spec, linked feedback and checklist index.

  4. Track as you go

    add_note for decisions, add_commit for hashes, and link_pr the moment a pull request exists.

  5. Complete with evidence

    get_completion_checklist, then complete_task with a confirmation for every item. Or abandon_task with a handoff summary.

  6. Let the loop close itself

    Completed public tasks auto-draft a changelog entry; publish_changelog notifies subscribers and voters.

Structured output schemas include task lifecycle messages, execution dates, summaries, review instructions, project details and feedback submitters; dependency project IDs and creation dates; and changelog dates and linked items. start_task and get_project_context share a context index schema with tags, appliesTo and exposed-page titles. Context sections include version and verification metadata; generated routing sections declare a string version, generated/readOnly flags and routeCount. Database reads include relationLabels, and row-value updates return updatedAt for conflict-safe edits.

Tasks

list_tasksTasks sorted by weighted demand score, filterable by status, type or label.
start_taskLock a task and receive its spec, linked feedback and project rules.
get_taskCompact by default (plan, current progress, subtasks, PRs, unresolved decisions, completion receipt) — request include sections like timeline or set detailMode: "full" for the legacy payload.
link_prLink the pull request that delivers the task — the primary deliverable, updated through merge.
add_noteRecord a decision or observation, or open/resolve a structured blocker or open question by kind.
update_taskEdit task fields, or replace the single currentProgress snapshot shown in compact retrieval.
add_api_exampleAttach a request/response example for an added or changed endpoint, with an optional before-response; credentials are redacted and re-sending a key replaces it.
add_screenshotAttach a screenshot of a changed page per state, theme, locale and viewport — inline base64 or a one-time upload URL for curl, with an optional before image.
complete_taskMark complete with evidence for every checklist item and a compact completion receipt; reports missing review evidence for the declared change kinds (or blocks when the project requires it) and releases the lock.
create_task delete_task add_commit list_review_artifacts remove_review_artifact create_subtask update_subtask delete_subtask add_task_dependency remove_task_dependency review_task abandon_task list_self_assessments

Context & graph

get_project_contextFull context summary — sections, critical rules, metadata. First call of every session.
resolve_contextMap touched file paths to governing sections, rules, risk domains and checklist scope.
resolve_context_graphBounded shadow-mode graph traversal — ranked candidates with full provenance.
validate_context_routingRun route fixtures against the production resolver before you rely on it.
get_context_graph compare_context_resolution list_context_sections get_context_section update_context_section patch_context_section delete_context_section verify_context_section sync_project_context get_section_blocks get_section_block_content list_context_routes upsert_context_route delete_context_route preview_context_route_migration render_context_routes_markdown list_context_relations upsert_context_relation delete_context_relation list_context_bundles get_context_bundle upsert_context_bundle delete_context_bundle list_context_fixtures upsert_context_fixture delete_context_fixture audit_context_health list_context_maintenance_candidates refresh_context_maintenance_candidates convert_context_maintenance_candidate preview_context_update report_context_feedback upsert_context_block

Repository-assisted context refresh

For a first setup or a substantial merged feature, prepare a reviewable context draft locally. Repomix creates the source packet locally and sends it to the selected agent or provider; Damper receives only the proposed sections and trusted preparation metadata.

MCP context responses may include optional freshness guidance: a missing receipt can suggest setup, while known managed edits or deletions can suggest the read-only CLI check and a reviewed refresh. Healthy receipts stay silent. MCP cannot inspect your checkout or source, so run the check locally; it never publishes automatically.

  1. Prepare locally

    damper context prepare --agent claude --repository-key default

  2. Review the draft

    Import the JSON draft in AI Context and inspect before/after content, tags, appliesTo and critical rules. Omitted sections remain unchanged.

  3. Publish explicitly

    Publish only after review. The server checks the project, repository key, section versions and receipt; stale drafts fail without a partial write. This workflow adds no MCP tools and never rewrites context automatically.

  4. Check freshness

    damper context check --repository-key default

Run the read-only check manually, on a schedule, or after the default branch receives a substantial merge. Use it as a downstream release/deployment gate; refresh the merged checkout, review and publish, then rerun the check. Pin the CLI version used by automation and keep publication as a separate reviewed step.

Pages & databases

The whole Pages workspace — with bounded reads, resumable polling and retry-safe writes for growing databases.

list_pages / search_pagesBrowse or full-text search every page, database and row in the project.
query_page_databaseFilter and project bounded row pages. With the default updatedAt ascending sort, save the returned resume cursor for incremental polls; in matches any supplied array member and contains requires all.
upsert_database_rowCreate or update one row by a canonical key unique within its parent database; unchanged retries are no-ops.
get_page_database_schemaRead every property ID, name, type and options, plus complete views and templates, in text and structured output without loading rows.
get_page_databaseRead the complete database and all row propertyValues keyed by property ID in text and structured output. Use for small databases or intentional full audits.
create_task_from_pageTurn a page into a roadmap task — links back and updates the row's roadmap property.

Database read tools include full JSON in text and structured output. get_page also includes row propertyValues keyed by property ID; query_page_database preserves the requested property/content projection.

Date values use YYYY-MM-DD strings, for example 2026-09-13; use null to clear them. Timestamp strings are reduced to their first ten characters without timezone conversion. For overdue rows, filter with { propertyId: reviewDateId, operator: "less_than", value: "2026-09-13" } and sort with { propertyId: reviewDateId, direction: "asc" }.

Sort by exactly one metadata field (createdAt or updatedAt) or a custom date propertyId. Empty dates sort last in both directions; row IDs break ties. Reuse cursors with the same sort. Only updatedAt ascending (the default) supports incremental change polling; cursors for other sorts are for pagination only.

get_page create_page update_page move_page duplicate_page archive_page get_page_backlinks create_database_property update_database_property update_database_row_values create_database_view update_database_view delete_database_view create_database_template

Feedback

list_feedback / get_feedbackItems with descriptions, sources, vote scores, AI summaries, occurrence counts, dates and linked tasks. Detail includes voter/comment dates and attachment URLs, with raw attachment data omitted. Seven days after an item is done or closed, its attachments, debug context and stored submitter email are deleted and its text and comments are redacted; visitor profiles are kept.
update_feedbackUpdate status or visibility. Returns id, title, status, type and isPublic, including when no changes are supplied. New feedback that contains personal data starts private; public pages always redact it.
reply_to_feedbackAnswer as a team member — the reply is visible to the customer.
link_feedback_to_taskConnect requests to work; submitters get notified when it ships.
report_issue

Releases

publish_changelogPublish a draft and notify subscribers of linked roadmap items.
list_changelogs create_changelog update_changelog add_to_changelog delete_changelog

Status page maintenance

list_status_maintenance_targetsDiscover group and monitor IDs, names and visibility for choosing affected components.
list_status_maintenanceRead maintenance windows, schedules, visibility and timeline updates, newest first.
create_status_maintenance / update_status_maintenanceSchedule maintenance or edit its details, scope and visibility. Dates use ISO 8601 UTC. Choose groupId or checkIds, never both. Omit both on creation for all monitors; omit both on edit to preserve scope, or set both to null for all monitors.
add_status_maintenance_updatePost progress and set state to scheduled, in_progress, completed or cancelled. Pass expectedState and expectedIsPublic from the latest read for conflict detection. Reuse a requestId UUID when retrying the same post.
edit_status_maintenance_updateCorrect a timeline message using maintenanceId and updateId. Historical state and visibility remain unchanged; no subscriber notification is sent.
delete_status_maintenancePermanently delete the window and timeline. Post a cancelled update instead to retain history.

All tools require projectId and follow API-key project access. Creation and timeline posts default to public; posts also default to notifying subscribers. Use isPublic: false for private work and notifySubscribers: false to suppress timeline notifications. Public window creation, edits and opted-in timeline posts queue durable subscriber notifications with background retries. A successful save confirms that the change and notification queue are committed, not that email has been delivered. Private windows still suppress monitoring alerts, and private updates cannot change the state of public maintenance.

{
  "name": "create_status_maintenance",
  "arguments": {
    "projectId": "YOUR_PROJECT_ID",
    "title": "Database upgrade",
    "scheduledStartAt": "2030-01-01T10:00:00Z",
    "scheduledEndAt": "2030-01-01T11:00:00Z",
    "isPublic": false
  }
}

Templates & modules

list_templates get_template update_template sync_templates list_modules get_module update_module sync_modules delete_module

Projects & workflow

list_projects returns accessible project IDs, names, slugs, ISO createdAt timestamps, and task/feedback counts. The get_project_context index includes an optional title for exposed pages. Both tools declare this metadata in their structured output schemas.

prepare_work / prepare_completionDeterministic work packets on the way in; drift and scope checks on the way out.
get_completion_checklistDetailed checklist branches to satisfy before completing a task.
open_public_pageThe correct public URL for roadmap, feedback, changelog or a roadmap item — never guess URLs.
list_projects get_agent_instructions get_project_settings update_project_settings patch_completion_checklist preview_completion_checklist_migration

Point your agent at it.

One URL, one key, a hundred tools. Your roadmap is already the backlog.

Get an API key Back to docs