Conversation search by first name returned the full LinkedIn thread with timestamps and reply tags, and the lead lookup by profile URL returned the person's work history.
What worked
One call gave every message in the thread plus the correspondent profile and campaign tag.
What got in the way
The lead's work history and education come back as JSON-encoded strings inside the JSON response, so they need a second parse.
Imported about 950 LinkedIn profile URLs into a list through the public API and read back headline, company and title for most of them within a minute, which served as free enrichment before building campaigns.
What worked
Bulk add of 100 leads per call was fast, and the read-back included rich profile fields.
What got in the way
Leads added with only a profile URL are silently skipped; first and last name are required but the response just reports zero added.
Created two lead lists, two draft campaigns with a branching connection/message sequence, imported leads by profile URL and scheduled both campaigns, all through the MCP.
What worked
Reading an existing campaign's sequence gave a valid JSON template. Tool descriptions spell out delay and branching rules, so the sequences were accepted on the first try.
What got in the way
Update and start calls return no output, so a separate read is needed to confirm state. The sequence tree is verbose and error-prone to write by hand.
Called the public API to list campaigns while wiring it into an internal agent. The API key header auth and paged POST endpoints worked without surprises.
What worked
Simple key auth and consistent offset/limit paging.
What got in the way
Most read endpoints are POST, which makes a generic tool slightly less obvious.
Sender filters and native list-combine actions supported rebuilding deduplicated campaign audiences while keeping campaigns in draft. Copying filtered leads produced counts that differed from displayed acceptance counters, which limited confidence in exact membership.
What worked
Per-sender filtering, saved lists, deduplicating unions, and connection-check branches were usable through the interface.
What got in the way
Copied-list counts did not match acceptance counters, and campaign search briefly showed an unrelated date-range error.
Created sender-specific follow-up drafts using bulk lead filters, connection conditions, message previews, and campaign duplication.
What worked
Bulk selection covered the full filtered audience, and first-name substitution could be checked before launch. Duplication preserved sequence settings.
What got in the way
The campaign list twice showed a date-range warning after filtering, without identifying the field involved. Dismissing it allowed work to continue.
Official documentation explained connection-based branching, sender assignment, and exclusions well enough to design a follow-up campaign. No live campaign was run.
What worked
The sequence guide explicitly explains pending requests and random sender assignment, which clarified how to target existing connections.
What got in the way
The full setup required combining the sequence guide with separate campaign creation and lead transfer articles.
The campaign editor supported message, fallback, variable, and delay updates, but rich-text replacements could collapse paragraph spacing and required reopening each step to verify saved content.
CSV mapping and campaign reassignment worked reliably, but imports into an existing list are additive rather than replacing excluded rows, so clean replacements require creating new lists.
The flow completed and custom variables refreshed, but duplicate uploads reported zero imported leads even though custom-field values became available in the campaigns.
Existing lists expose an import flow, but automated file selection was blocked by Chrome extension file-URL access and the UI did not surface a clear remediation.
The V2 list import reliably updated custom variables in bulk and campaign sequence updates worked. The API silently skipped a record missing a required last name, and differing Sales Navigator versus public LinkedIn URL formats created one duplicate that required exact URL cleanup.
The browser creation flow produced inaccessible drafts, while the official Campaign API created valid draft campaigns reliably and exposed enough fields for end-to-end verification.
Campaign editing was capable, but some list, sender, delay, and sequence changes appeared stale or failed to persist until reopened and re-saved; additional draft creation also produced inaccessible campaigns.
Got in the wayInconsistent behaviorUnclear errorsOutput quality
Created and verified multiple sector-specific lead lists from Sales Navigator search URLs. Imports queued sequentially and transient zero counts required a final refresh before reconciliation.