Tool catalog

Tool	Scope	Effect
`get_connection`	`workspaces:read`	Identify the connected Emailx actor, environment, key grants and expiry.
`list_workspaces`	`workspaces:read`	Discover currently accessible customer or agency client workspaces.
`get_workspace_capabilities`	`workspaces:read`	Request-level permissions and role, without granting execution authority.
`list_members`	`team:read`	One teammate page and a separate owner; internal proxy identities excluded.
`get_reply_routing`	`team:read`	Channel-specific notification modes and owner options; no changes.
`list_senders`	`senders:read`	Stored account metadata and local blockers, without live provider checks.
`get_sender`	`senders:read`	Authorized channel-specific sender details; no credentials.
`find_contacts`	`contacts:read`	Find contacts by exact case-insensitive email, or list a bounded page in an explicit workspace.
`select_people`	`contacts:read`	Preview local eligibility for explicit contact IDs or allowlisted filters; read-only, requires an idempotency key and does not enroll contacts.
`list_campaigns`	`campaigns:read`	List saved campaigns in one explicit workspace and channel.
`get_campaign`	`campaigns:read`	Read a campaign and its saved steps before changing it.
`get_campaign_configuration`	`campaigns:read`	Saved setup and revision; does not approve content.
`configure_campaign`	`campaigns:read`, `campaigns:configure`	Inactive setup with revision protection; no activation by this tool.
`create_campaign_draft`	`campaigns:write`	Create an inactive draft for human review.
`pause_campaign`	`campaigns:read`, `campaigns:write`	Pause a specifically requested campaign after reading its current state.
`list_conversations`	`inbox:read`	Find saved campaign conversations with replies in a specific workspace/channel.
`get_conversation`	`inbox:read`	Read saved conversation messages.
`get_reply_context`	`inbox:read`	Read up to 40 saved messages with freshness, completeness, sender readiness, and the revision required to reply; never refreshes provider history.
`send_inbox_reply`	`inbox:send`	Queue one explicitly confirmed reply using the original sender and thread; requires exact current revision and idempotency key. Unknown outcomes must be reconciled, never automatically resent.
`get_workspace_report`	`analytics:read`	Read all-time saved delivery totals for one workspace/channel.
`list_templates`	`templates:read`	List saved regular and smart email template summaries.
`get_template`	`templates:read`	Read saved template fields and the revision used for editing.
`create_template`	`templates:write`	Save regular or smart template settings; does not generate or send email.
`update_template`	`templates:write`	Save requested template edits with revision protection.
`import_linkedin_profiles`	`campaigns:import`, `campaigns:read`	Queue lookup for 1 to 100 explicit profile URLs. Requires the default-off execution feature and can consume provider quota; no launch or send.
`get_operation`	`operations:read`, `campaigns:read`	Read durable import or LinkedIn activation progress and safe results. Requires the execution feature.
`list_operation_rows`	`operations:read`, `campaigns:read`	Read paginated outcomes for one import. Requires the execution feature.
`cancel_operation`	`operations:write`, `campaigns:read`, original command grant	Requires `campaigns:import` for imports or `campaigns:launch` for activations; cancel only when the server confirms it is safe and does not undo in-flight work.
`review_linkedin_campaign`	`campaigns:launch`	Create a short-lived local readiness review. Does not verify provider health, quota or delivery, and does not activate. Requires the execution feature.
`launch_linkedin_campaign`	`campaigns:launch`	Explicitly activate a reviewed manual LinkedIn campaign for native scheduled dispatch. Requires `confirm_launch=true` and the execution feature; no immediate batch or delivery verification.
`review_linkedin_resume`	`campaigns:launch`	Create a short-lived local review for resuming a paused, previously launched manual LinkedIn campaign. Reports uncertain attempts and other blockers; does not resume. Requires the execution feature.
`resume_linkedin_campaign`	`campaigns:launch`	Explicitly resume a reviewed paused LinkedIn campaign for the periodic native scheduler. Requires `confirm_resume=true` and the execution feature; keeps saved progress and never retries uncertain attempts.
`get_linkedin_campaign_progress`	`campaigns:read`	Read locally saved status counts and timestamps; saved activity does not prove delivery.
`get_linkedin_prospect`	`campaigns:read`	Read one prospect’s saved progress under an explicit campaign; no provider lookup.

Inputs that matter

Every workspace operation requires workspace_id. Campaigns, inbox and reporting require channel: email or linkedin. Lists default to 25 records and accept limit from 1 to 100 plus an opaque cursor. All 18 write tools require an idempotency_key. An email draft accepts a name only. LinkedIn drafts support audience and existing step fields. pause_campaign reads current state first and cannot recall in-flight activity. The LinkedIn URL import accepts explicit HTTPS LinkedIn profile URLs for an eligible inactive manual draft; it is not a broad search and can use provider lookup quota. People import separately previews CSV/XLSX and saves contacts without enrollment. Email preparation, credit-confirmed generation, exact-message approval, reviewed launch, and confirmed inbox replies are separate actions. Email launch schedules within configured windows; it does not verify delivery. Inbox context may be incomplete or stale and does not refresh provider history. Unknown reply outcomes require reading the original operation; never retry automatically. See email execution, inbox replies, People import, and LinkedIn review and launch. Resume, reply drafts, deletion and broad inbox history refresh remain unavailable.

Typed results

Successful calls return typed data and meta, including request ID, HTTP status and replay status. Lists retain pagination. Errors have isError=true and a structured error; they are never simulated success. Results are capped at 512 KiB, and unexpected API fields fail contract validation rather than leaking through. Download the machine-readable tool catalog for exact schemas. Tool annotations describe effects; they do not replace a user’s authorization or live API permission checks.

Connect an assistant

Workflows and access

Inputs that matter

Typed results

​Inputs that matter

​Typed results

Inputs that matter

Typed results