Servidor MCP
Permita que clientes de IA descubram e usem as mesmas capacidades governadas do Blocchi.
O Blocchi expõe MCP em POST /api/mcp usando JSON-RPC 2.0. A revisão do próprio servidor é 2026-07-28, e ele também negocia 2025-06-18 e 2025-03-26, de modo que clientes padrão se conectam sem tratamento especial.
Discovery
Clientes padrão usam o handshake initialize comum e não precisam de cabeçalhos específicos do Blocchi:
A resposta devolve a protocolVersion negociada. Se a revisão solicitada não for suportada, o servidor responde com a sua própria.
server/discover é a alternativa nativa mais completa do Blocchi: informa também as extensões suportadas e onde ficam os metadados de autorização.
Após o handshake, as chamadas fixam a revisão negociada em MCP-Protocol-Version. Mcp-Method e Mcp-Name são dicas de roteamento opcionais; quando enviadas, precisam concordar com o corpo JSON-RPC.
Uma requisição não autenticada sempre recebe 401 com o desafio WWW-Authenticate indicando os metadados do recurso protegido, para que o cliente descubra o servidor de autorização antes de ter credenciais.
Semântica de governança
MCP não contorna autorização, prontidão, políticas de fonte nem ações governadas do Blocchi. Ferramentas com efeito de mutação podem retornar input_required, um pedido de confirmação e um estado opaco vinculado ao ator, tenant, operação e input exatos.
O modelo não recebe credenciais de provedores simplesmente porque está chamando via MCP.
Catálogo de ferramentas ao vivo
A lista abaixo é gerada a partir dos mesmos registries MCP usados pelo servidor, em vez de manter uma segunda lista manual.
Protocolo atual
2026-07-28| Ferramenta | O que faz |
|---|---|
blocchi.acknowledge_alert | Mark one authorized alert as acknowledged or unacknowledged for the current actor without changing the source business event. |
blocchi.add_opportunity_offering | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.add_outbound_prospect | Create or reuse an account/contact and place the person into an outbound experiment with ICP reasoning and a next-message angle. This does not contact the prospect. |
blocchi.approve_outbound_draft | Human-governed approval of the exact rendered outbound message. Approval still does not send it. |
blocchi.archive_customer | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.archive_opportunity | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.archive_person | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.archive_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.begin_operating_run | Start or resume an idempotent run for the authenticated Automation Identity. Use this before a recurring external-agent routine manages Growth + Revenue work. |
blocchi.claim_human_work | CAS-claim one assigned Human Work item before responding. |
blocchi.close_opportunity_lost | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.close_opportunity_won | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. Preview and execution enforce configured evidence-backed win readiness for the current semantic stage. |
blocchi.complete_revenue_commitment | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.complete_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.confirm_revenue_fact | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.create_alert_rule | Create a governed personal alert, or an organization alert when the actor is Tenant Admin. Personal metric alerts retain the actor's current governed analytics scope. |
blocchi.create_customer | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.create_opportunity | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.create_outbound_experiment | Create a controlled outbound learning experiment. This does not send messages. |
blocchi.create_person | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.create_report | Create a governed semantic report when the caller has report-author capability and the target product surface is entitled. |
blocchi.create_revenue_commitment | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.create_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.decide_advisory_recommendation | Accept, reject, or request changes on the exact recommendation revision/current-state token through the same Human Work approval ledger used by the UI. |
blocchi.decide_human_work | Submit an accountable human decision. Consequential approvals resume the same immutable workflow run/revision/node and require the exact captured state token; stale state is invalidated. |
blocchi.delete_alert_rule | Delete a visible personal rule or, for Tenant Admin, an organization rule. This changes alert configuration but not historical events/alerts. |
blocchi.delete_subscription | Unsubscribe the current actor from one delivery subscription without changing the underlying alert rule or report. |
blocchi.describe_api | Teach an AI client how to use Blocchi's REST API, including auth, readiness, canonical IDs and common workflows. Use this before inventing endpoints or provider-specific queries. |
blocchi.describe_data_plus | Return Data+ implementation maturity, entitlement/access state and safe operational readiness information. |
blocchi.evaluate_opportunity_readiness | Evaluate evidence-backed opportunity readiness through the same engine enforced by semantic Revenue actions. purpose=advance requires targetStageId; purpose=win evaluates close-won readiness for the current semantic stage and omits targetStageId. |
blocchi.execute_advisory_recommendation | Apply one accepted recommendation as ordered registered governed actions. Each action re-previews current state and enforces its own customer authorization/approval policy; partial success is explicit. |
blocchi.explore_data | Browse the Data Explorer business/source view with server-derived viewer scope, safe search/filter/sort and admin-only sanitized raw payload. |
blocchi.find_customers | Find customers in the current actor's governed Revenue scope, including source provenance, readiness, intelligence and next action. |
blocchi.find_opportunities | Find opportunities in the current actor's governed Revenue scope with commercial-model, segment/offering, risk, readiness and next-action filters. |
blocchi.find_people | Find people in the current actor's governed Revenue scope with customer relationships, next action, readiness and source provenance. |
blocchi.find_revenue_interactions | Find governed interactions only within the actor's linked customer, person or opportunity scope. |
blocchi.finish_operating_run | Persist the routine checkpoint and completion/waiting state so the next scheduled run can continue safely without blind retries. |
blocchi.get_activation_state | Determine whether this tenant can receive trustworthy Blocchi value now and return the minimum governed next step. Use this before inventing setup steps or treating unavailable data as zero. |
blocchi.get_advisory_recommendation | Read one customer recommendation and, when reviewable, resolve its exact current-state governed-action diff. Customer admin authorization remains authoritative. |
blocchi.get_communication_analysis_context | Read one linked interaction's bounded retained content for evidence analysis. Output is evidence context only; extracted facts, commitments and next actions must use existing governed proposal/action tools. |
blocchi.get_ecosystem_coverage | Show owned, connected, degraded and missing business capabilities across the customer lifecycle. |
blocchi.get_entity | Get one canonical Blocchi entity with roles and source-record provenance. Provider-native fields are intentionally not exposed as the contract. |
blocchi.get_finding | Get a governed finding and the evidence currently supporting it. |
blocchi.get_finding_evidence | Return only the structured evidence behind a governed finding. |
blocchi.get_first_revenue_brief | Return the evidence-bounded first-value Revenue brief derived from the tenant's governed Commercial Model, Brand, source/data, Revenue Plan/Plays and Growth state. Missing evidence remains explicit rather than being invented. |
blocchi.get_human_work | Show one assigned Human Work item with exact workflow/state binding when applicable. |
blocchi.get_operating_brief | Return Automation Identity authority, last successful checkpoint, lifecycle gaps and bounded external context. Then use existing Revenue/Growth read tools only within the identity's configured authority. |
blocchi.get_outbound_engagement | Get the full account/contact/experiment/activity context for one prospect. Optionally include a uniquely attributed assessment URL. |
blocchi.get_report | Read one tenant-authored semantic report definition and visualization contract. |
blocchi.get_report_catalog | Get governed semantic entities, measures and dimensions that may be used to author reports. Provider-native schemas are not the contract. |
blocchi.get_revenue_fact_registry | List the effective typed commercial fact registry. Methodologies are templates over these keys, not hard-coded truth schemas. |
blocchi.inspect_whatsapp_business | Read connected WhatsApp Business phone health and capability state without exposing provider credentials. |
blocchi.list_advisory_configuration_actions | List registered Blocchi-only configuration actions that typed Advisory recommendations are allowed to compose. This is a catalog read, not an execution capability. |
blocchi.list_advisory_recommendations | List privacy-safe Advisory operating-model recommendations addressed to the current customer tenant. This does not expose consultant-only evidence. |
blocchi.list_alert_rules | List organization rules plus personal rules visible to the current actor. |
blocchi.list_alerts | List materialized governed alerts addressed to the current actor, optionally filtered by product, severity or acknowledgement state. |
blocchi.list_ecosystem_integrations | List Blocchi's provider-neutral business capability catalog. Provider-private MCP tool names and credentials are never returned. |
blocchi.list_events | List the governed cross-product activity feed the current actor is authorized to see. Events retain readiness, evidence context and deep links. |
blocchi.list_evidence | List governed observed/reported/documented/derived evidence for an allowed subject. Missing evidence stays unknown. |
blocchi.list_external_context | Read bounded external-system context with source, freshness and authority metadata instead of copying external systems of record into Blocchi. |
blocchi.list_findings | List governed findings the caller may see. The first implementation exposes Data Quality findings; future Revenue/Legal findings join the same contract. |
blocchi.list_human_work | List accountable decisions, reviews and information requests assigned to the current human user, role or governed queue. |
blocchi.list_interview_purposes | List optional governed interview purposes and their downstream uses for the tenant's entitled business applications. |
blocchi.list_metrics | List governed Blocchi metric/indicator definitions without exposing provider-native fields. |
blocchi.list_my_communication_accounts | List only the current human user's connected Google Workspace or Microsoft 365 accounts, capability state and synchronization health. Provider tokens are never exposed. |
blocchi.list_my_communication_reviews | List commercially relevant email/calendar items that Blocchi refused to attach automatically because the Revenue link is ambiguous. |
blocchi.list_my_overdue_revenue_tasks | List open or in-progress Revenue tasks in the current actor's governed scope whose due timestamp is in the past. |
blocchi.list_operator_actions | List the curated Revenue, Growth and paid-media actions available to this Automation Identity after capability, risk, source, data-scope and emergency-stop policy are applied. |
blocchi.list_outbound_experiments | List the tenant's founder-led outbound experiments and their explicit persona/problem/message/offer hypotheses. |
blocchi.list_outbound_queue | Return prioritized outbound work for agents or humans: new prospects, drafts, follow-ups, replies and completed assessments that need action. |
blocchi.list_reports | List predefined and tenant-authored reports for a governed Revenue, Legal or shared surface. |
blocchi.list_revenue_commitments | List first-class commercial promises with source, due state, confidence and completion independently from Tasks. |
blocchi.list_revenue_facts | List commercial facts with status, confidence, source and evidence under the caller's viewer scope and sensitivity rules. |
blocchi.list_revenue_tasks | List the current actor's governed Revenue tasks. Use owner/status filters for personal or team work queues. |
blocchi.list_revenue_timeline | Read a bounded, deduplicated customer/person/opportunity activity timeline without rendering raw provider HTML. |
blocchi.list_sources | List tenant source datasets, persistence mode and last sync information. |
blocchi.list_subscriptions | List only the current actor's alert/event/report delivery subscriptions and preferences. |
blocchi.list_surfaces | List available Blocchi Revenue, Legal and Advisory surfaces. |
blocchi.list_whatsapp_identity_reviews | Human-user view of unresolved or reviewed inbound WhatsApp phone identities. This tool never creates or links CRM records. |
blocchi.list_whatsapp_templates | List synchronized WhatsApp Business templates and their current provider approval status. |
blocchi.log_outbound_activity | Append an outbound touch or downstream signal after it happens. For touch_logged, store the exact message that was actually sent. |
blocchi.log_revenue_interaction | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.move_opportunity_stage | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. Preview and execution enforce configured evidence-backed stage readiness. |
blocchi.override_opportunity_readiness | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. This high-risk action is human-decision only and works only for override-enabled profiles. purpose=advance requires targetStageId; purpose=win omits it. |
blocchi.preview_connected_email | Preview the exact From account, recipient, subject/body binding, reply target and attachment hashes before any connected-mailbox send. |
blocchi.preview_report | Execute an unsaved Report Builder definition through the governed analytics engine. Caller viewer scope is always enforced. |
blocchi.propose_revenue_fact | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.propose_revenue_interaction_evidence | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. Submit only bounded evidence extracted from an authorized Interaction; facts/promises remain inferred/proposed and include model provenance. |
blocchi.query_metric | Compute/query a metric through Blocchi's governed semantic engine. Always inspect readiness: blocked/unknown is not zero. |
blocchi.reassign_human_work | Reassign one Human Work item through the same CAS and role policy used by the product UI and REST API. |
blocchi.reassign_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.reject_revenue_fact | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.reopen_opportunity | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.request_ecosystem_action_approval | Create idempotent Human Work for a restricted external business action. The request is bound to the exact governed preview and assigned to authorized human approvers. |
blocchi.request_operator_action_approval | Create Human Work bound to the exact run, action input and governed preview when an internal Revenue/Growth action cannot execute autonomously. |
blocchi.reschedule_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.respond_to_interview | Submit one text or audio-transcript answer to a governed optional interview. Reported statements are preserved as reported evidence and are not silently treated as truth. |
blocchi.revenue_today | Return the current human actor's deterministic Revenue Today projection, including reason and priority-factor codes that clients can localize. |
blocchi.review_communication_link | Explicitly link or ignore one ambiguous communication item. This uses the same durable source decision as Human Work. |
blocchi.run_ecosystem_action | Two-step business-capability action. First call previews a reviewed capability on an approved connection. Execute only by repeating the business input with the returned previewBinding and an idempotencyKey. For non-autonomous actions include an approved Human Work ID. Raw remote MCP server/tool identifiers are not accepted. |
blocchi.run_operator_action | Two-step curated internal action. First call previews the task-oriented Revenue/Growth action. Execute only by repeating the exact input with previewBinding, idempotencyKey and confirm=true; include approved Human Work when required. |
blocchi.run_report | Execute a saved authored report through the governed analytics engine with the caller's current viewer scope. |
blocchi.save_outbound_draft | Store the exact rendered message proposed for a prospect. Saving a draft never sends it. |
blocchi.save_subscription | Create or update the current actor's delivery preference for an authorized alert rule, event stream or report. Supports in-app/email and immediate or digest delivery. |
blocchi.search_entities | Search organizations, people and other shared-core canonical entities within the caller's tenant. |
blocchi.send_connected_email | Send a previously previewed exact one-to-one or transactional email after explicit human confirmation. Requires the preview binding and an idempotency key. |
blocchi.send_whatsapp | Two-step semantic send. First call returns an exact policy/provider preview. To execute, call again with the same business input plus previewBinding, idempotencyKey and confirm=true. Recipient comes from a Revenue person, never a raw phone number. |
blocchi.set_opportunity_next_action | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.set_opportunity_stakeholder | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.set_outbound_next_action | Set the recommended human/agent follow-up and due time for a prospect. |
blocchi.show_customer | Show one governed customer operating view with relationships, opportunities, tasks, timeline, intelligence and source provenance. |
blocchi.show_opportunity | Show one opportunity directly with customer, stage history, stakeholders, offerings, tasks, timeline, source and readiness. |
blocchi.show_person | Show one governed Revenue person with customer relationships, tasks, timeline and source provenance. |
blocchi.show_revenue_interaction | Show one governed Revenue interaction including source provenance and privacy retention class. |
blocchi.start_interview | Start an optional purpose-driven interview. Blocchi governs which facts the interview is allowed to learn. |
blocchi.summarize_outbound_experiment | Measure response, assessment-completion and qualified-conversation signals for one message hypothesis. |
blocchi.sync_my_communication_account | Run bounded durable cursor catch-up for one connected communication account owned by the current human user. |
blocchi.trigger_sync | Consequential tenant-admin action. The server uses MCP multi-round-trip input to obtain explicit user confirmation for the exact dataset before executing the sync. |
blocchi.update_alert_rule | Update a visible personal rule or, for Tenant Admin, an organization rule. Authorization and governed metric capability are revalidated. |
blocchi.update_customer | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.update_opportunity | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.update_person | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.update_report | Version and update an existing tenant-authored semantic report. Changing its product surface is intentionally not supported in v1. |
blocchi.update_revenue_commitment | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.update_revenue_task | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. Completion, rescheduling and reassignment use their dedicated tools. |
blocchi.upsert_customer_person_relationship | First call returns an exact governed preview. To execute, call again with the same business input plus previewBinding and idempotencyKey; set confirm=true only after accepting an approval-required preview. |
blocchi.verify_advisory_recommendation | Record post-application outcome evidence without treating configuration application as causal business success. |
Última atualização em