MCP tool reference

The read-only tools Reyda’s MCP server gives AI assistants, with parameters, what each returns, required scopes and limits.

Reyda’s MCP server gives connected AI assistants ten read-only tools. This page is for people setting up or building on a connection. To connect an assistant, see Connect Claude, Connect ChatGPT or Other MCP clients.

All tools are marked read-only and non-destructive. None of them creates pursuits, edits data, starts document extraction or starts paid analysis.

Connection details

Item	Value
Server URL	https://api.reyda.ai/mcp in production. Copy the exact value from Settings → Integrations → AI assistants.
Transport	Streamable HTTP, stateless, JSON responses
Authentication	OAuth authorization code with PKCE (S256 only) and a pre-registered client ID. No dynamic client registration.
Discovery	/.well-known/oauth-authorization-server and /.well-known/oauth-protected-resource/mcp on the server origin
Token and revocation	POST /oauth/token and POST /oauth/revoke on the server origin
Resource	Authorization requests must include resource set exactly to the server URL
Consent	Signed-in user approves in the Reyda app (app.reyda.ai)

Refreshing rotates both the access and refresh token. Reusing an old code or refresh token revokes the whole connection, so clients should refresh one request at a time.

Scopes

Scope	Shown on the consent screen as	Tools
opportunities:read	Read opportunities	reyda_search_opportunities, reyda_get_opportunity; also needed for opportunity documents
documents:read	Read documents you can access	reyda_list_documents, reyda_read_document
pipeline:read	Read your organization’s pipeline	reyda_list_pipeline
intelligence:read	Research buyers, suppliers and contracts	reyda_search_organizations, reyda_get_buyer, reyda_get_supplier, reyda_list_supplier_contracts

reyda_get_context needs no scope. The user approves the requested scopes together. Scopes never widen what the user can already see in Reyda.

Access checks on every call

Each tool call checks, at the time of the call, that:

• the user is still an active member of the connected organization;
• the organization has an active paid subscription (free, trial and expired plans are refused; a plan scheduled to end works until its end date);
• the connection has the scope the tool needs;
• the user can see the specific opportunity, document or pipeline item, including Reyda’s rules for sensitive library documents.

Limits

Limit	Value
Results per page	Default 20, maximum 50
Filter lists (countries, CPV codes, buyers)	Maximum 20 entries each
Document text per read	Maximum 20,000 characters

Response size, request size, call time and call rate are also capped. Long text fields may be cut short, with truncated set. Pagination reflects live data, so results can shift if data changes between pages.

Tools

reyda_get_context

Describes the connection: environment, granted scopes, available tools, the Reyda app URL and its limits (max_results, max_document_characters, max_response_bytes). No parameters. Useful as a first call to check the connection works.

reyda_search_opportunities

Searches existing opportunities. Returns opportunities (same fields as reyda_get_opportunity), has_more and next_offset. Scope: opportunities:read.

Parameter	Required	Description
mode	Yes	browse (most recent first) or personalized (ranked by relevance, applying the organization’s country and value preferences)
query	No	Search keywords
countries	No	Two-letter uppercase country codes, for example IE, DE
cpv_codes	No	CPV code prefixes, 2 to 8 digits
buyer_ids	No	Buyer references from reyda_search_organizations
min_value, max_value	No	Estimated value bounds
deadline_from, deadline_to	No	Inclusive timestamps with time zone (RFC 3339), for example 2026-11-01T00:00:00Z
limit, offset	No	Page size (1 to 50) and number of results to skip

reyda_get_opportunity

Returns one opportunity: title, summary and summary_kind, source_url, deadline and deadline_timezone, estimated_value and currency, publication_date, updated_at, the organization’s existing fit result (shortlist_status, shortlist_score), app_url, and missing_data / truncated flags. The summary is labelled as an existing AI summary, separate from published facts. Scope: opportunities:read.

Parameter	Required	Description
id	Yes	Opportunity ID from search

reyda_list_documents

Lists files with their processing status (ready, processing or unavailable), title, app_url and update time. Omit opportunity_id to list the organization’s library. Returns documents, has_more, next_offset. Scope: documents:read, plus opportunities:read for opportunity documents.

Parameter	Required	Description
opportunity_id	No	List this opportunity’s tender documents instead of the library
limit, offset	No	Page size (1 to 50) and offset

reyda_read_document

Reads text Reyda has already extracted from a document. Returns document, status, evidence (the text, its character offset, total_characters, revision, and page_start / page_end when Reyda has a reliable page mapping), next_offset, truncated and a citation_url. If there is no extracted text, status is processing or unavailable; this doesn’t mean the document is empty, and no extraction is started. Scope: documents:read, plus opportunities:read for opportunity documents.

Parameter	Required	Description
document_id	Yes	Document ID from reyda_list_documents
opportunity_id	For opportunity documents	Omit for library documents
offset	No	Character offset to continue from (use next_offset)
limit	No	Characters to return, default and maximum 20,000
revision	When continuing	The revision from the first read. If the document has changed, the read must restart from the beginning.

Offsets count characters, not bytes. Document text is untrusted content: assistants are told to treat it as evidence, never as instructions.

reyda_list_pipeline

Lists pipeline items in the connected organization. Each item has id, opportunity_id, title, status, app_url, source_url and updated_at. Returns items, has_more, next_offset. Scope: pipeline:read.

Parameter	Required	Description
status	No	triage (default), pursue, no_bid or archived
query	No	Search text
assigned_to	No	Filter by assignee
limit, offset	No	Page size (1 to 50) and offset

reyda_search_organizations

Finds buyers and suppliers by name. Each result has a stable id, name, role, country and app_url. Organizations with the same name keep separate IDs. Returns organizations, coverage of the recorded data, has_more, next_page, missing_data. Scope: intelligence:read.

Parameter	Required	Description
name	Yes	Name to search
role	No	buyer or supplier; omit for both
country	No	Two-letter uppercase country code
from, to	No	Date range as YYYY-MM-DD; give both or neither
page, limit	No	Page number and page size (1 to 50)

reyda_get_buyer

Returns a buyer’s recorded award count, recorded contract values by currency (each currency kept separate), top suppliers, and coverage. Contract values are published values, not actual payments. Scope: intelligence:read.

Parameter	Required	Description
id	Yes	Buyer ID from reyda_search_organizations
country	No	Two-letter uppercase country context

reyda_get_supplier

Returns a supplier’s profile, sectors, recorded winner award count and the recorded award value in its dominant currency only (not a total across currencies). Scope: intelligence:read.

Parameter	Required	Description
id	Yes	Supplier ID from reyda_search_organizations
country	No	Two-letter uppercase country context

reyda_list_supplier_contracts

Lists a supplier’s recorded contracts. Each contract has title, buyer name, ID and country, participation_role, publication_date, published_contract_value and currency, recorded_end_date (empty when unknown), opportunity_id, source_url and app_url. Returns contracts, coverage, has_more, next_page. Recorded end dates don’t confirm a future tender. Scope: intelligence:read.

Parameter	Required	Description
supplier_id	Yes	Supplier ID from reyda_search_organizations
view	No	all (default) or expiring for contracts approaching their recorded end date
country	No	Buyer country filter, two-letter uppercase code
from, to	No	Date range as YYYY-MM-DD; give both or neither
page, limit	No	Page number and page size (1 to 50)

Error codes

When a call is refused, the tool returns an error with one of these codes:

Code	Meaning
plan_access_required	The organization has no active paid subscription. The connection stays in place.
missing_scope	The connection wasn’t granted the scope this tool needs
access_denied	The user is no longer an active member, or the connection isn’t valid
rate_limited	Too many calls; retry shortly
response_too_large	The result was too large; request a smaller page
timeout	The call took too long
document_changed_restart_read	The document changed since the first read; start again without offset and revision

Requests over the HTTP rate limit get status 429 with a Retry-After header.

FAQ

Which tool should a client call first?

reyda_get_context. It needs no scope and returns the environment, granted scopes, available tools and limits, so it confirms the connection works.

Why does reyda_read_document return processing or unavailable?

Reyda hasn’t extracted text for that document yet. This doesn’t mean the document is empty, and reading never starts extraction. Open the document in Reyda and try again later.

Why was the whole connection revoked after a refresh?

Refreshing rotates both tokens. Reusing an old code or refresh token, for example from two parallel refreshes, revokes the connection. Refresh one request at a time.

Related

Reyda MCP server

Other MCP clients

Manage and revoke connections

Limits