Skip to content

Agentic access via MCP (Model Context Protocol)

POST
/mcp
curl --request POST \
--url https://api.partners.collaterate.com/v1/mcp \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '{}'

A single JSON-RPC 2.0 endpoint speaking MCP’s Streamable HTTP transport. Every one of this API’s 11 business operations (see the rest of this document) is exposed as an MCP tool – tools/list returns only the tools your credential’s granted scopes allow, and tools/call runs that tool’s own REST operation’s logic directly, in-process, rather than proxying the call over HTTP, so the request/response contract for any given tool is the SAME contract its REST operation documents here. This path’s own body is a JSON-RPC envelope ({"jsonrpc":"2.0","id":...,"method":"tools/list"|"tools/call",...}), not modeled as a typed request/response schema here since JSON-RPC’s shape is protocol-defined, not this API’s own contract to describe field-by-field the way every other path is.

Media typeapplication/json

A JSON-RPC 2.0 request. See the Model Context Protocol specification for the full envelope shape.

object
Examplegenerated
{}

A JSON-RPC 2.0 response (a result or a protocol-level error).

Media typeapplication/json
object
Examplegenerated
{}

Missing, expired, or otherwise invalid credential. Not retryable as-is.

Produced by the API gateway’s authorizer, before the request reaches any application code, so every cause – no Authorization header, a malformed one, a token that is expired, wrongly signed, from the wrong pool, or belongs to a credential that has been disabled – yields this identical response. The gateway is configured to answer in the same application/problem+json shape as everything else, so you do not need a second parser for this status; what it cannot do is tell you which of those causes applied.

Media typeapplication/problem+json

RFC 9457 application/problem+json body. code is the published, stable, machine-readable field to branch your integration logic on – detail is a human-readable string that may be reworded over time and must not be parsed.

This table is the whole published set: every code this API can return appears below, and nothing below is unreachable. test/docs/openapi-matches-reality.test.ts compares the enum to the codes the handlers actually construct, so a code added to one side and not the other fails the build rather than shipping.

Published code values:

code HTTP status Meaning Retry?
invalid_cursor 400 The cursor query parameter could not be decoded. No – restart pagination with no cursor. Do not resend the same value.
invalid_site_id 400 The siteId filter is not a positive integer, or it names a site outside your grant. detail says which. No – fix the parameter against GET /v1/me’s grantedSiteIds.
invalid_order_number 400 The order number in the path is not a positive integer. No – fix the path.
invalid_product_id 400 The productId path segment on GET /v1/products/{productId} does not match the SLO_/SLS_-prefixed shape. No – fix the path.
invalid_service_type 400 GET /v1/products’s serviceType filter is not one of the values this API can order (STOCK or POD). No – fix the parameter.
updated_since_not_supported 400 An updatedSince parameter was sent. Rejected rather than ignored, on purpose – see “No updatedSince / no change polling” above. No – remove the parameter; there is no change-polling mechanism to switch to.
invalid_request 400 POST /v1/orders’s request body failed validation – malformed JSON, a missing or malformed field, an unknown key, or a body-level rule such as partnerLineId uniqueness or a misplaced shipToor GET /v1/orders/submissions’s status/limit filter is malformed. detail names the exact violation. No – fix the request.
invalid_submission_id 400 The submissionId path segment on GET /v1/orders/submissions/{submissionId} is not a well-formed UUID. No – fix the path.
invalid_quantity 400 The quantity on POST /v1/products/{productId}/quote is missing or not a positive integer. No – fix the request body.
invalid_token 401 The credential is missing, expired, or invalid. No – mint a new token, retry once, then investigate the credential.
insufficient_scope 403 The token lacks the scope required for this operation. No – mint a token with the required scope.
order_not_found 404 No such order, or one that exists outside your granted sites (see above). No.
product_not_found 404 No such product, or one that exists but belongs to a site outside your granted sites (same rule as order_not_found). No.
quote_product_not_found 404 The product resolved in our own catalog but Collaterate could not price it. Rare. No.
submission_not_found 404 No submission exists with this id for your credential – whether it never existed or belongs to another partner, identically. No.
ordering_not_provisioned 409 POST /v1/orders’s siteId has no ordering user configured. This is set up per (partner, site) by staff at onboarding; a partner cannot provision it themselves. No – ask Partner Integrations to provision an ordering user for the site, then retry.
partner_order_id_reused 409 POST /v1/orders’s partnerOrderId was already used, with a request body that does not match this one. Resending the identical body under the same partnerOrderId is safe and returns the original submission (see the operation description); reusing it for a materially different order is rejected instead of silently creating a second order. No – use a new partnerOrderId, or resend the exact original body to poll the existing submission.
quote_invalid 422 Collaterate rejected the quote on business-validation grounds (e.g. quantity over the product’s configured maximum). detail carries Collaterate’s own validation message. No – fix the request body (e.g. reduce quantity).
rate_limit_exceeded 429 Rate limit or daily quota exceeded. Yes – honor Retry-After.
internal_error 500 Unexpected server-side failure. Detail is always the fixed string below, never the underlying error. No, or with caution – if persistent, contact Partner Integrations.
service_unavailable 503 Transient failure – typically the database path, but also a Lambda that crashed, timed out, or hit its own concurrency limit before any handler ran. Yes, with backoff.

Three statuses are answered by the API gateway before your request reaches application code: 401 (authorizer denial), 429 (usage-plan throttle), and 503 (the Lambda integration itself failed). All three are configured to return this same document with the codes above, so one parser and one branch on code covers every error this API produces. The 403 on GET /v1/ping is the sole exception anywhere in the API – it comes from the edge firewall, which is not ours to shape, and carries no problem document.

object
type
required

Always the literal string about:blank today; reserved by RFC 9457 for future use.

string
Allowed value: about:blank
title
required

Short, human-readable summary of the HTTP status (e.g. “Not Found”).

string
status
required

The HTTP status code, repeated in the body for convenience.

integer
code
required

Stable, machine-readable error identifier. Safe to branch on. See the table above.

string
Allowed values: invalid_cursor invalid_site_id invalid_order_number invalid_product_id invalid_quantity invalid_service_type updated_since_not_supported invalid_request invalid_submission_id invalid_token insufficient_scope order_not_found product_not_found quote_product_not_found submission_not_found ordering_not_provisioned partner_order_id_reused quote_invalid rate_limit_exceeded internal_error service_unavailable
detail
required

Human-readable explanation. Do not parse this – it may be reworded without notice.

string
requestId
required

Echoes the request’s id. Include this when contacting Partner Integrations about a specific failed call – it resolves directly to one log entry.

string
Example
{
"type": "about:blank",
"title": "Unauthorized",
"status": 401,
"code": "invalid_token",
"detail": "The credential presented with this request is missing, expired, or invalid.",
"requestId": "8f3c1e2a-..."
}

Your rate limit or daily quota was exceeded. Retryable – honor the Retry-After header (seconds) before your next attempt.

Like the 401, this comes from the gateway’s usage plan rather than from application code, and is configured to carry the same problem document as every other error so a single parser covers the whole API.

Media typeapplication/problem+json

RFC 9457 application/problem+json body. code is the published, stable, machine-readable field to branch your integration logic on – detail is a human-readable string that may be reworded over time and must not be parsed.

This table is the whole published set: every code this API can return appears below, and nothing below is unreachable. test/docs/openapi-matches-reality.test.ts compares the enum to the codes the handlers actually construct, so a code added to one side and not the other fails the build rather than shipping.

Published code values:

code HTTP status Meaning Retry?
invalid_cursor 400 The cursor query parameter could not be decoded. No – restart pagination with no cursor. Do not resend the same value.
invalid_site_id 400 The siteId filter is not a positive integer, or it names a site outside your grant. detail says which. No – fix the parameter against GET /v1/me’s grantedSiteIds.
invalid_order_number 400 The order number in the path is not a positive integer. No – fix the path.
invalid_product_id 400 The productId path segment on GET /v1/products/{productId} does not match the SLO_/SLS_-prefixed shape. No – fix the path.
invalid_service_type 400 GET /v1/products’s serviceType filter is not one of the values this API can order (STOCK or POD). No – fix the parameter.
updated_since_not_supported 400 An updatedSince parameter was sent. Rejected rather than ignored, on purpose – see “No updatedSince / no change polling” above. No – remove the parameter; there is no change-polling mechanism to switch to.
invalid_request 400 POST /v1/orders’s request body failed validation – malformed JSON, a missing or malformed field, an unknown key, or a body-level rule such as partnerLineId uniqueness or a misplaced shipToor GET /v1/orders/submissions’s status/limit filter is malformed. detail names the exact violation. No – fix the request.
invalid_submission_id 400 The submissionId path segment on GET /v1/orders/submissions/{submissionId} is not a well-formed UUID. No – fix the path.
invalid_quantity 400 The quantity on POST /v1/products/{productId}/quote is missing or not a positive integer. No – fix the request body.
invalid_token 401 The credential is missing, expired, or invalid. No – mint a new token, retry once, then investigate the credential.
insufficient_scope 403 The token lacks the scope required for this operation. No – mint a token with the required scope.
order_not_found 404 No such order, or one that exists outside your granted sites (see above). No.
product_not_found 404 No such product, or one that exists but belongs to a site outside your granted sites (same rule as order_not_found). No.
quote_product_not_found 404 The product resolved in our own catalog but Collaterate could not price it. Rare. No.
submission_not_found 404 No submission exists with this id for your credential – whether it never existed or belongs to another partner, identically. No.
ordering_not_provisioned 409 POST /v1/orders’s siteId has no ordering user configured. This is set up per (partner, site) by staff at onboarding; a partner cannot provision it themselves. No – ask Partner Integrations to provision an ordering user for the site, then retry.
partner_order_id_reused 409 POST /v1/orders’s partnerOrderId was already used, with a request body that does not match this one. Resending the identical body under the same partnerOrderId is safe and returns the original submission (see the operation description); reusing it for a materially different order is rejected instead of silently creating a second order. No – use a new partnerOrderId, or resend the exact original body to poll the existing submission.
quote_invalid 422 Collaterate rejected the quote on business-validation grounds (e.g. quantity over the product’s configured maximum). detail carries Collaterate’s own validation message. No – fix the request body (e.g. reduce quantity).
rate_limit_exceeded 429 Rate limit or daily quota exceeded. Yes – honor Retry-After.
internal_error 500 Unexpected server-side failure. Detail is always the fixed string below, never the underlying error. No, or with caution – if persistent, contact Partner Integrations.
service_unavailable 503 Transient failure – typically the database path, but also a Lambda that crashed, timed out, or hit its own concurrency limit before any handler ran. Yes, with backoff.

Three statuses are answered by the API gateway before your request reaches application code: 401 (authorizer denial), 429 (usage-plan throttle), and 503 (the Lambda integration itself failed). All three are configured to return this same document with the codes above, so one parser and one branch on code covers every error this API produces. The 403 on GET /v1/ping is the sole exception anywhere in the API – it comes from the edge firewall, which is not ours to shape, and carries no problem document.

object
type
required

Always the literal string about:blank today; reserved by RFC 9457 for future use.

string
Allowed value: about:blank
title
required

Short, human-readable summary of the HTTP status (e.g. “Not Found”).

string
status
required

The HTTP status code, repeated in the body for convenience.

integer
code
required

Stable, machine-readable error identifier. Safe to branch on. See the table above.

string
Allowed values: invalid_cursor invalid_site_id invalid_order_number invalid_product_id invalid_quantity invalid_service_type updated_since_not_supported invalid_request invalid_submission_id invalid_token insufficient_scope order_not_found product_not_found quote_product_not_found submission_not_found ordering_not_provisioned partner_order_id_reused quote_invalid rate_limit_exceeded internal_error service_unavailable
detail
required

Human-readable explanation. Do not parse this – it may be reworded without notice.

string
requestId
required

Echoes the request’s id. Include this when contacting Partner Integrations about a specific failed call – it resolves directly to one log entry.

string
Example
{
"type": "about:blank",
"title": "Too Many Requests",
"status": 429,
"code": "rate_limit_exceeded",
"detail": "The request rate limit has been exceeded. Retry after the indicated delay.",
"requestId": "8f3c1e2a-..."
}
Retry-After
integer

Seconds to wait before retrying.