Use the Video History Project with your AI assistant
Search the collection, read records, and collect citations using your own AI assistant. These tools provide access to the public material available on this site.
Connect your assistant
In an assistant that supports remote MCP servers, add a server named Video History Project and paste this address. Choose Streamable HTTP if asked for a transport. No account or API key is required.
After connecting, ask your assistant to describe the collection. It can discover six tools:
describe_collection, search_records, get_record,
list_terms, get_related_records, and export_citations.
Use your browser assistant
A compatible browser assistant can discover tools while you visit this site. Ask it to search, read the current page, or retrieve a citation. Searches open the results on the page so you can follow along.
Browser tool availability has not been checked. The ordinary site works without browser tools.
WebMCP is experimental and browser support varies. API availability does not mean an assistant is connected.
Try a research question
- Find material about video synthesizers published between 1970 and 1980.
- Find writing by this author and give me the citations.
- Read this record and show me its linked source documents.
What the tools can read
Tools read existing public record text, abstracts, summaries, bibliography and document links. Generated summaries are identified separately. Scans are linked; these tools do not read or OCR them. Related records are suggestions, not evidence of a historical relationship.
Choose whether to cite the VHP page or the publication it describes. Missing bibliographic details remain missing. Public corrections are identified as reader contributions. See the site terms; linked documents may have different rights.
Read the collection and citation guide · Collection information as JSON
Connection help
If an assistant cannot connect, check that it supports remote Streamable HTTP MCP and use the address above. Opening that address in a normal tab is not a connection test. A login prompt is not required by this server. If browser tools are unavailable, use the normal search or connect through MCP. If requests are limited, wait the indicated time before retrying.
For developers
Public JSON uses the same research service as MCP and WebMCP. Responses include canonical URLs, explicit pagination and source labels. GET responses support ETag revalidation.
- /api/v1/search: q, type, author, subject, from, to, page, limit.
/api/v1/records/{id}: section, offset, limit, version./api/v1/resolve?url=…: resolve a VHP ID or historical URL; the same text parameters apply.- /api/v1/terms: field (author or subject), q, page, limit.
/api/v1/records/{id}/related: limit./api/v1/citations?ids=1,2&format=ris&target=publication: up to 25 IDs; format text or ris; target page or publication./api/v1/documents/{id}: destinations and discoverable citing records for a document landing page.
Search defaults to 25 results over JSON and 10 over MCP; the maximum is 25. Follow nextPage with the same filters and limit. Record text defaults to the body and at most 12,000 UTF-16 code units. Other sections are listed with their source and size. Follow content.nextOffset with the same section and returned version. A record_changed response means restart reading. Responses never silently substitute a summary for body text.
Tools are read-only. The shared request budget is 600 research requests per minute; a limited request returns HTTP 429 and Retry-After. Query errors explain what to correct. A missing record returns 404; a withdrawn record returns 410.
Verified September 17, 2026 with the MCP C# and Python SDK clients (2.2.0), and Chromium 151 with experimental WebMCP enabled. Browser checks cover search navigation, pagination, record text, citations and the ordinary mobile site with WebMCP disabled. Connection instructions use standard MCP settings.