PRODUCTS / AI OUTREACH
Multi-agent outbound that reads each lead, drafts in voice, and respects the limits your accounts impose. Five surfaces, one reasoning trace.
INTEGRATIONS
Connection requests, DMs, InMail.
Cold + warm, threaded.
Business API.
X
DMs to verified leads.
DMs to business accounts.
One agent stack. Every surface above. LinkedIn first by volume; the four others run the same reasoning.
HOW IT WORKS
LinkedIn, Email, WhatsApp, Instagram, X. OAuth or paste. Tokens encrypted at rest.
CSV upload. Per-row, per-channel validation. The agent doesn't touch a row that doesn't pass.
Six archetypes. Edit any prompt. Pick the reasoning depth per campaign — shallow, standard, deep.
30/day cap. 14-day warmup. Circuit breakers. Watch the agent work in real time — and stop when the numbers say so.
THE WORKBENCH
Three surfaces your team works in every day. Rank the queue. Read the agent. Hold the limits. Every decision audit-logged, every prompt editable, every cap visible.
DRAFTS
11 / 30 today
Mira Sato
Head of Growth · Bowtie
Daniel Park
VP RevOps · Loom
Aiko Watanabe
Founder · Maple
Tomás Lima
Head of Sales · Hex
SEQUENCE · 4 OF 30 SENT TODAY
PROMPT
For SHALLOW: profile, role, company.
For STANDARD: + posts, shared interests.
For DEEP: + articles, mutuals, 3-5 sources.
Never fabricate.
TOOLS · 9
AGENT · MASTRA · openai('gpt-5.4')
DAILY CAP
30 / 30
WARMUP
d14 / 14
WARMUP RAMP
14 days
ABOVE CAP → BLOCKED · NO OVERRIDE
OPERATIONS
Above cap → blocked. Hard limit, not a target.
Volume rises 3 → 30 across two weeks. No skip.
Provider unstable → the agent stops. Auto-recovers when provider is healthy.
Three operational commitments — cap, warmup, breaker — enforced in code, not policy. The numbers above ran on Carlos's pilot campaigns; the LinkedIn account that produced them is still alive at week 16. Cohort baselines published with their data at GA.
SECURITY
Every message, every reasoning step, every override — written, signed, exportable.
Role-based access. LinkedIn tokens isolated from message bodies isolated from lead lists.
EU and US-only regional deployment. No data crosses borders without your config.
Lead-list PII redacted in agent logs and human-review surfaces. Tokens AES-256-GCM at rest.
WHY THIS
AI-FIRST
Research, drafting, and reply-analysis as separate, prompt-editable agents. Every step audit-logged. Open box, not closed.
OPERATIONS
30/day cap, 14-day warmup, working-hours window — and a provider-level circuit breaker that trips when sends start to misfire. Above the line the agent refuses; the rep can't override.
REACH
LinkedIn, Email, WhatsApp, X, Instagram. Same agent stack runs each. Multi-channel out of the box, not on the roadmap.
BUILT BY
Shippit's engineers ship production AI for B2B and fintech. AI Outreach runs on the same engineering bench, the same operational discipline, and the same security posture we deliver to clients with real money on the line.
Engagement
Regulated tokenization for Brazilian securities. Vendors quoted 12–16 months; we shipped in 4.
Engagement
AI infra protocol. Marketing site, API-key dashboard, headless CMS — shipped in 8 weeks.
Engagement
AI mentorship at company scale. 30+ Mastra agents, 120+ mentors, AnaFly coaching in production.
EARLY ACCESS
One operator per week. Cohort seats lock in founder-direct Slack access, a 30-minute architecture walkthrough, and pre-GA pricing that won't apply at launch.