mcp.placeholder.com

MCP server

Everything on this site as tools for your AI agent: image, photo, avatar, QR, barcode and brand URLs, synthetic people, the mock API, the Brightfjord company with its eval questions, fixtures with answer keys, dummy files, HTTP test endpoints and safe test values. Remote, free, read-only, no sign-up.

Connect

Endpoint: https://mcp.placeholder.com/mcp. Streamable HTTP, stateless, JSON responses, no authentication. Opening it in a browser gives a 405 because MCP clients talk to it with POST; the info page has the snippets below for more clients.

Claude Code

claude mcp add --transport http placeholder https://mcp.placeholder.com/mcp

Claude desktop and claude.ai

Settings > Connectors > Add custom connector > URL: https://mcp.placeholder.com/mcp

Cursor (`.cursor/mcp.json`)

{"mcpServers": {"placeholder": {"url": "https://mcp.placeholder.com/mcp"}}}

VS Code (`.vscode/mcp.json`)

{"servers": {"placeholder": {"type": "http", "url": "https://mcp.placeholder.com/mcp"}}}

Windsurf, Gemini CLI, stdio-only clients

{"mcpServers": {"placeholder": {"serverUrl": "https://mcp.placeholder.com/mcp"}}}   // Windsurf
{"mcpServers": {"placeholder": {"httpUrl": "https://mcp.placeholder.com/mcp"}}}     // Gemini CLI
npx mcp-remote https://mcp.placeholder.com/mcp                                      // stdio bridge

Try it with curl

curl -s https://mcp.placeholder.com/mcp \
  -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Tools

24 tools, all read-only and deterministic: the same arguments give the same result. Arguments marked * are required.

URL builders (fetch nothing)

Tool	What it does	Arguments
`image_url`	A classic placeholder image URL, plus a ready `<img>` tag.	`width`*, `height`, `bg`, `fg`, `text`, `format`, `scale`
`photo_url`	A photo URL matched to a subject and cropped to size.	`subject`*, `width`, `height`, `seed`, `format`, `grayscale`, `blur`, `scale`
`avatar_url`	An avatar URL: initials, shapes, identicon or silhouette, never a face.	`name`*, `size`, `style`, `shape`, `bg`, `fg`, `format`, `seed`, `scale`
`qr_url`	A QR code image URL.	`text`*, `size`, `ecc`, `margin`, `fg`, `bg`, `format`
`barcode_url`	A barcode URL; without a value, a test-safe code. Can also list test-safe values.	`type`, `value`, `seed`, `prefix`, `check`, `scale`, `height`, `human_readable`, `format`, `test_codes`
`brand_assets`	Logo, mark, favicon, social card, palette sheet and `tokens.css` URLs for a fictional brand.	`brand`*, `theme`, `format`

People, test values and text

Tool	What it does	Arguments
`get_person`	One synthetic persona with test-range IDs, phone and IBAN.	`country`, `pid`
`generate_people`	A batch of personas, plus CSV/JSON/NDJSON/SQL download URLs for up to 10,000 rows.	`country`, `count`, `seed`, `fields`, `bulk_count`
`safe_test_values`	Safe test values with sources; omit the category for the list.	`category`, `country`, `limit`
`lorem_text`	Lorem ipsum and its permanent URL.	`type`, `count`, `seed`, `start`, `format`
`localized_text`	Realistic copy in 17 languages, parallel across languages per seed.	`lang`*, `type`, `count`, `seed`, `format`
`stress_text`	Strings that break layouts and encodings, with length metadata.	`kind`, `count`, `seed`
`pseudo_localize`	Pseudo-localized text to catch hard-coded strings and truncation.	`text`*, `expand`, `brackets`

Mock API and company in a box

Tool	What it does	Arguments
`mock_api`	Query the mock REST API; the returned URL also accepts simulated writes.	`resource`*, `id`, `filters`, `q`, `sort`, `order`, `page`, `limit`, `embed`, `expand`, `country`
`company_overview`	Profile, collections, fields, wiki pages and URL patterns of Brightfjord. Start here.	`company`, `version`
`company_query`	Records of one collection with filters, text search and paging.	`collection`*, `filters`, `q`, `limit`, `page`, `company`, `version`
`company_record`	One record by ID; threads and channels include their messages.	`collection`, `id`, `company`, `version`
`company_wiki_page`	A wiki page as Markdown; omit the page to list them.	`page`, `company`, `version`
`company_eval_questions`	Eval questions with exact answers and evidence record IDs.	`limit`, `page`, `kind`, `include_answers`, `company`, `version`

Brands, fixtures, files and HTTP

Tool	What it does	Arguments
`brand_kit`	A full brand kit: palette with contrast, fonts, logo URLs, voice and copy. Omit the brand for the curated list.	`brand`
`list_fixtures`	Fixtures with answer keys; omit the kind for the catalog.	`kind`, `subkind`, `tag`, `language`, `variant`
`get_fixture`	A fixture's manifest (file URLs, sha256) and optionally its answer key.	`id`*, `kind`, `include_answers`
`dummy_files`	Dummy files filtered by category, extension, size or breakage.	`category`, `extension`, `max_bytes`, `min_bytes`, `search`, `broken`, `limit`
`http_test_endpoints`	HTTP test endpoints with example URLs.	`group`

Tips for agents

Results link to the public https:// URLs on via, data and files. For bulk data (CSV, SQL, ZIP bundles, 10,000 people) fetch those links rather than making many tool calls.
For agent and RAG evals, start with company_overview, then company_eval_questions to grade answers.
Long results are trimmed at about 24,000 characters, and say so; page or filter to get the rest.

Server info

https://mcp.placeholder.com/info page with connect snippets
https://mcp.placeholder.com/index.jsonthe same as JSON
https://mcp.placeholder.com/server.jsonMCP registry metadata (com.placeholder/mcp)

Nothing about requests is logged. Everything the tools return is synthetic and deterministic, like the rest of the service.