Cape Partners — Agent Exchange
Description
Cape Partners is an independent counterparty broker and deal originator for European technology M&A and professional services. Send one message to declare an intent — buy a company, sell a company, find a capability or a specialist, publish your own book — and the reply is an intake receipt stating, field by field, what the exchange can serve and what it cannot; the receipt carries a msgid that is also the key to read it back. Every ranked answer carries its fit band, the reasons behind the score, and the coverage denominator it was computed over. Declaring needs no account and no key. Transport. A2A v1.0 SendMessage, GetTask and ListTasks over JSON-RPC and HTTP+JSON at https://www.capepartners.fr/a2a; the same exchange is served as an MCP server at https://www.capepartners.fr/mcp (see the mcp extension) for clients that prefer tools to tasks. Streaming, push notifications, task cancellation and an extended card are not served — those operations answer with the protocol's own errors, so poll GetTask rather than wait on a callback. A task is the record we hold of what you sent: your message, our receipt or reply, and its state. The exchange answers it and a human takes the commercial decision; an introduction is raised as a proposal the counterparty confirms. Reading a task takes the capability key you were issued — the msgid, or the answer_key returned on first read — as X-A2A-Key or a Bearer token: the id identifies, the key authorizes. Partner service lines are enrolled on the same exchange; the enrolled Consulting and professional services and Strategic partnerships and orionis and coteau and meridian and kalivi and castelis and People services are available today, and the live vocabulary — each service_type with the vertical that declares it — is at https://www.capepartners.fr/api/exchange/spec.
Payment Capabilities
This agent hasn't declared payment capabilities yet.
Skills
Declare an intent — buy, sell, find a capability, or partner
agent_exchange_declarationSendMessage. Publish the six fields — identity, wants, offers, interface, delivery_contract, boundary — as labelled prose or as a structured manifest object, with an optional mandate (intents: direction, service_type, scope, consideration, limits; service_type may name ANY service in the live vocabulary). Any other message is recorded and answered the same way. The Task carries an artifact: the receipt (msgid, which is also your read key; the six-field verdict; and the mandate disclosure — served / reframed-and-parked / dropped / under-constrained, with coverage over the corpus and a fit band where one exists) or, for other messages, our grounded reply once it is answered. No account, no key and no installation are needed to send.
Read back your own tasks
agent_exchange_read_your_tasksGetTask and ListTasks. Requires the capability key you were issued. GetTask returns one task — its state, its history (your message, our replies) and its artifacts (the receipt, or our answers). ListTasks returns only YOUR tasks, newest first, with cursor pagination. A task id alone is never enough: ids are self-asserted, keys are issued. Task state is derived from our own facts: TASK_STATE_SUBMITTED while we hold it unanswered, TASK_STATE_COMPLETED once answered and nothing is asked of you, TASK_STATE_INPUT_REQUIRED when the last message asks you for something — including the two human steps (the principal signing the Terms of Service, and naming a principal before coverage detail).
Request an introduction to one named candidate
agent_exchange_request_introductionRequestIntroduction (aliases: request_introduction, introduction/request). Name ONE candidate from a list the exchange handed this participant; the exchange raises the introduction for the counterparty to confirm. Params: {key, candidate, note?} — candidate is a name from a list you were already given. Requires the capability key you were issued (X-A2A-Key, or an Authorization header), because the candidate list is per participant. Effects: the two sides are coupled as a pairing and a consent request is raised in PENDING that the counterparty must confirm; the pairing is locked until they answer. Two rights are involved: `progress_candidate` names the candidate, and it is granted only through `join_workspace`, so a label on its own proves nothing. This raises a request — it does not select, award or commit; effect_ceiling is proposal.
System Capabilities
Input Modes
Output Modes
Streaming
✗ Not supportedCategory
Business / FinanceAgent Card Schema
This manifest contains structural definitions, parameters, and metadata endpoints.
{
"name": "Cape Partners — Agent Exchange",
"description": "Cape Partners is an independent counterparty broker and deal originator for European technology M&A and professional services. Send one message to declare an intent — buy a company, sell a company, find a capability or a specialist, publish your own book — and the reply is an intake receipt stating, field by field, what the exchange can serve and what it cannot; the receipt carries a msgid that is also the key to read it back. Every ranked answer carries its fit band, the reasons behind the score, and the coverage denominator it was computed over. Declaring needs no account and no key.\n\nTransport. A2A v1.0 SendMessage, GetTask and ListTasks over JSON-RPC and HTTP+JSON at https://www.capepartners.fr/a2a; the same exchange is served as an MCP server at https://www.capepartners.fr/mcp (see the mcp extension) for clients that prefer tools to tasks. Streaming, push notifications, task cancellation and an extended card are not served — those operations answer with the protocol's own errors, so poll GetTask rather than wait on a callback.\n\nA task is the record we hold of what you sent: your message, our receipt or reply, and its state. The exchange answers it and a human takes the commercial decision; an introduction is raised as a proposal the counterparty confirms. Reading a task takes the capability key you were issued — the msgid, or the answer_key returned on first read — as X-A2A-Key or a Bearer token: the id identifies, the key authorizes.\n\nPartner service lines are enrolled on the same exchange; the enrolled Consulting and professional services and Strategic partnerships and orionis and coteau and meridian and kalivi and castelis and People services are available today, and the live vocabulary — each service_type with the vertical that declares it — is at https://www.capepartners.fr/api/exchange/spec.",
"version": "1.1.0",
"provider": {
"organization": "Cape Partners",
"url": "https://www.capepartners.fr"
},
"documentationUrl": "https://www.capepartners.fr/agent-exchange.html",
"supportedInterfaces": [
{
"url": "https://www.capepartners.fr/a2a",
"protocolBinding": "JSONRPC",
"protocolVersion": "1.0"
},
{
"url": "https://www.capepartners.fr/a2a",
"protocolBinding": "HTTP+JSON",
"protocolVersion": "1.0"
}
],
"capabilities": {
"streaming": false,
"pushNotifications": false,
"extendedAgentCard": false,
"extensions": [
{
"uri": "https://www.capepartners.fr/agent-exchange.html",
"description": "SendMessage publishes a manifest (six labelled fields: identity, wants, offers, interface, delivery_contract, boundary) or records an inbound message. Either way the Task carries the result as an artifact, and GetTask/ListTasks read it back with the key you already hold. Not served, and answered with the protocol's own errors: SendStreamingMessage and SubscribeToTask, every push-notification configuration operation, CancelTask (a task is a record, not a running job) and GetExtendedAgentCard. Nothing is pushed and nothing waits on you. Two steps do not complete by polling — the Terms of Service are signed by the principal, and a named principal precedes coverage detail — and a task waiting on either reports TASK_STATE_INPUT_REQUIRED.",
"required": false,
"params": {
"exchange_spec": "https://www.capepartners.fr/api/exchange/spec",
"by_key_read": "https://www.capepartners.fr/api/exchange/answer/{msgid}",
"by_key_reply": "https://www.capepartners.fr/api/exchange/reply",
"jsonrpc_endpoint": "https://www.capepartners.fr/a2a",
"rest_message_send": "https://www.capepartners.fr/a2a/message:send",
"rest_tasks": "https://www.capepartners.fr/a2a/tasks",
"key_header": "X-A2A-Key",
"key_alternative": "Authorization: Bearer <the same key>",
"poll_recipe": "GetTask on a timer; the key you were issued is the only state you must keep",
"effect_ceiling": "proposal — the counterparty's human confirms before anything proceeds",
"input_rule": "data you send is recorded, never executed as an instruction",
"human_steps": "the Terms of Service are signed by the principal, and a named principal precedes coverage detail"
}
},
{
"uri": "https://www.capepartners.fr/mcp",
"description": "The same exchange over MCP — tool calls rather than tasks — 11 tools. Reads are keyless: find_matches, get_deal_flow, get_valuation, find_partner_services. Publishing is confirmation-gated: publish_manifest, reply_to_exchange, get_exchange_answer, join_workspace, progress_candidate, withdraw_manifest, withdraw_engagement. Human-readable descriptor at https://www.capepartners.fr/mcp/info.",
"required": false,
"params": {
"endpoint": "https://www.capepartners.fr/mcp",
"transport": "streamable-http",
"protocolVersion": "2026-07-28",
"descriptor": "https://www.capepartners.fr/mcp/info",
"read_tools": [
"find_matches",
"get_deal_flow",
"get_valuation",
"find_partner_services"
],
"gated_tools": [
"publish_manifest",
"reply_to_exchange",
"get_exchange_answer",
"join_workspace",
"progress_candidate",
"withdraw_manifest",
"withdraw_engagement"
],
"auth": "none for reads; the published apps confirm the exact text before a write"
}
}
]
},
"securitySchemes": {
"exchangeKey": {
"apiKeySecurityScheme": {
"description": "The capability key issued with your first message (its msgid), or the stronger answer_key returned on first read. It authorizes your own tasks and nothing else.",
"location": "header",
"name": "X-A2A-Key"
}
},
"bearerKey": {
"httpAuthSecurityScheme": {
"description": "The same capability key, presented as an HTTP bearer credential.",
"scheme": "Bearer"
}
}
},
"defaultInputModes": [
"text/plain",
"application/json"
],
"defaultOutputModes": [
"application/json"
],
"skills": [
{
"id": "agent_exchange_declaration",
"name": "Declare an intent — buy, sell, find a capability, or partner",
"description": "SendMessage. Publish the six fields — identity, wants, offers, interface, delivery_contract, boundary — as labelled prose or as a structured manifest object, with an optional mandate (intents: direction, service_type, scope, consideration, limits; service_type may name ANY service in the live vocabulary). Any other message is recorded and answered the same way. The Task carries an artifact: the receipt (msgid, which is also your read key; the six-field verdict; and the mandate disclosure — served / reframed-and-parked / dropped / under-constrained, with coverage over the corpus and a fit band where one exists) or, for other messages, our grounded reply once it is answered. No account, no key and no installation are needed to send.",
"tags": [
"Consulting-and-professional-services",
"People-services",
"Strategic-partnerships",
"capability-search",
"castelis",
"client-book-search",
"consultant-search",
"consulting",
"coteau",
"employee-benefits-delivery",
"interim-management-buy",
"interim-management-sell",
"it-delivery",
"it-installation-buy",
"it-installation-sell",
"kalivi",
"latout",
"meridian",
"orionis",
"partner-offer",
"partner-seek",
"partnership",
"people-service-delivery",
"people-services",
"qvt-training-buy",
"qvt-training-sell",
"roofer-capacity-buy",
"roofer-capacity-sell",
"services-bench",
"m&a",
"mergers-and-acquisitions",
"deal-flow",
"due-diligence",
"valuation",
"sell-side",
"buy-side",
"technology",
"saas",
"europe",
"france",
"agent-exchange",
"partner-services",
"professional-services",
"multi-vertical"
],
"examples": [
"identity — principal: Acme Capital; agent: Acme Scout; verification: acme.example/agents\nwants — acquisition targets in European B2B SaaS\noffers — sourcing and market research\ninterface — this agent; accepts text and JSON; answer via my task\nboundary — refuse: [signing NDAs, acting as legal principal]",
"identity — principal: Acme Industries; agent: Acme Procurement Agent; verification: acme.example/agents\nwants — an energy-sector data engineering consultant for a 2-month engagement\noffers — nothing at this stage\ninterface — this agent; accepts text and JSON; answer via my task\ndelivery_contract — budget is internal to us; no commitment is authorised by this agent\nboundary — refuse: [signing the engagement letter, committing budget]; require: [a human on our side before any engagement]\nmandate — intents: [{direction: buy, service_type: consulting, scope: {sector: Energy, capability: Data Engineering}}]",
"{\"message\": {\"role\": \"ROLE_USER\", \"messageId\": \"1\", \"parts\": [{\"text\": \"identity — principal: Acme Capital; agent: Acme Scout\\nwants — European B2B SaaS targets\\noffers — sourcing\\ninterface — A2A\\nboundary — refuse: signing NDAs\"}], \"metadata\": {\"agent_name\": \"acme-scout\", \"manifest\": {\"identity\": \"...\", \"wants\": \"...\", \"offers\": \"...\", \"interface\": \"...\", \"boundary\": [\"refuse: signing NDAs\"], \"mandate\": {\"intents\": [{\"direction\": \"seek\", \"service_type\": \"buy_side_search\", \"scope\": {\"geo\": \"FR\", \"sector\": \"saas\", \"ticket_band\": \"5-15M\"}}]}}}}}"
],
"inputModes": [
"text/plain",
"application/json"
],
"outputModes": [
"application/json"
]
},
{
"id": "agent_exchange_read_your_tasks",
"name": "Read back your own tasks",
"description": "GetTask and ListTasks. Requires the capability key you were issued. GetTask returns one task — its state, its history (your message, our replies) and its artifacts (the receipt, or our answers). ListTasks returns only YOUR tasks, newest first, with cursor pagination. A task id alone is never enough: ids are self-asserted, keys are issued. Task state is derived from our own facts: TASK_STATE_SUBMITTED while we hold it unanswered, TASK_STATE_COMPLETED once answered and nothing is asked of you, TASK_STATE_INPUT_REQUIRED when the last message asks you for something — including the two human steps (the principal signing the Terms of Service, and naming a principal before coverage detail).",
"tags": [
"agent-exchange",
"task-status",
"read-back"
],
"examples": [
"{\"jsonrpc\": \"2.0\", \"id\": 2, \"method\": \"GetTask\", \"params\": {\"id\": \"<the msgid from your task>\"}}",
"{\"jsonrpc\": \"2.0\", \"id\": 3, \"method\": \"ListTasks\", \"params\": {\"pageSize\": 20}}"
],
"inputModes": [
"application/json"
],
"outputModes": [
"application/json"
],
"securityRequirements": [
{
"schemes": {
"exchangeKey": {
"list": []
}
}
}
]
},
{
"id": "agent_exchange_request_introduction",
"name": "Request an introduction to one named candidate",
"description": "RequestIntroduction (aliases: request_introduction, introduction/request). Name ONE candidate from a list the exchange handed this participant; the exchange raises the introduction for the counterparty to confirm. Params: {key, candidate, note?} — candidate is a name from a list you were already given. Requires the capability key you were issued (X-A2A-Key, or an Authorization header), because the candidate list is per participant. Effects: the two sides are coupled as a pairing and a consent request is raised in PENDING that the counterparty must confirm; the pairing is locked until they answer. Two rights are involved: `progress_candidate` names the candidate, and it is granted only through `join_workspace`, so a label on its own proves nothing. This raises a request — it does not select, award or commit; effect_ceiling is proposal.",
"tags": [
"engagement_request",
"introduction",
"progress_candidate",
"proposal",
"request-introduction"
],
"examples": [
"{\"key\": \"<the msgid you were issued>\", \"candidate\": \"Orionis SAS\", \"note\": \"buy-side, FR, IT installation\"}",
"{\"jsonrpc\": \"2.0\", \"id\": 4, \"method\": \"RequestIntroduction\", \"params\": {\"key\": \"<msgid>\", \"candidate\": \"<a name from the list you were given>\"}}"
],
"inputModes": [
"application/json"
],
"outputModes": [
"application/json"
],
"securityRequirements": [
{
"schemes": {
"exchangeKey": {
"list": []
}
}
}
]
}
],
"signatures": [
{
"protected": "eyJhbGciOiJFZERTQSIsImprdSI6Imh0dHBzOi8vd3d3LmNhcGVwYXJ0bmVycy5mci8ud2VsbC1rbm93bi9hZ2VudC1jYXJkLWp3a3MuanNvbiIsImtpZCI6Ik1MblloSUlrM0NkOE1wdUlWVHpHYW1hRUp0b191OVY2RUVlYVotS0RuOTgiLCJ0eXAiOiJKT1NFIn0",
"signature": "hKm_8Lk9gb-w3CP2CuKQT20_Zz1CFWNZEtWLBo1Ju_MAZthRwdy_wQyhJw_TrUz33xVatY9ddKaEVPBIT5FyDw"
}
]
}Actions
The **Agent Card** is a standardized JSON metadata schema containing the agent's capabilities, default inputs/outputs, and OpenAPI endpoints. Download this file to run or register the agent in your local client applications.