Home/Products/AI Agents

AI agents · the agent library

A whole outbound team. On demand.

Specialized agents that do the work you'd otherwise hire for: enrichment, inbox triage, reply drafting, objection handling, meeting booking, intent monitoring. Compose them, observe them, trust them on your own timeline.

Eight specialized agents in the library, more shipping monthly
Three autonomy modes per agent: Suggest · Auto-under-cap · Full-auto
Drag-and-drop chain composer — branch on confidence, intent, stage
Full per-action trace + one-click rollback on any decision
Build your own (no code) — or bring your own LLM

Start free trial See the roster

7-day trial · Credit card required · Cancel anytime

PLATE · 01AI Agentsfig.

Agent roster · workspace6 active

Lead Enricher

Enriched 124 leads · 3m ago

Auto

Inbox Triager

Classified 38 replies · 5m ago

Auto

Reply Drafter

9 drafts awaiting review

Suggest

Objection Handler

Drafted 4 rebuttals · 12m ago

Suggest

Meeting Booker

Held 3 meetings · today

Auto

Intent Monitor

2 hot accounts flagged

Auto

2.1k

Actions/day

24/7

Uptime

11h

Saved/rep/wk

Why agents, not one SDR4 metrics

Specialists shipped, observed, and trusted on your terms.

METRIC / 01

Specialized agents

Each shipped, edited and ran independently.

METRIC / 02

2.1k

Actions/day

Median across a 6-rep workspace, last 30 days.

METRIC / 03

11h

Saved per rep/week

Median time returned to reps from agent automation.

METRIC / 04

Autonomy modes

Suggest · auto-under-cap · full-auto. Per agent.

How it works

Library → autonomy → workflows

Pick agents, decide how autonomous each one is, chain them into the work you want done.

STEP / 01

Pick from the library

Eight production agents are available the moment you sign in. Each is a specialist: enrichment, triage, drafting, objection handling, booking, intent monitoring, hygiene, follow-up pacing. Turn one on, turn them all on, or compose them into a chain.

STEP / 02

Set the autonomy you trust

Three modes per agent: Suggest (draft for you to approve), Auto-under-cap (act inside a daily budget without approval), Full auto (no cap). Promote an agent from Suggest to Auto once it's earned the trust — DitLead shows you when.

STEP / 03

Compose into workflows

Chain agents together. The Intent Monitor surfaces a hot account → the Enricher fills in firmographics → the Drafter writes a first touch → the Booker proposes calendar windows when the prospect bites. Every handoff is logged.

Inside the platform

Four design choices behind every agent

Most 'AI sales' products give you one closed agent. DitLead gives you a library, a chain, a trace, and a builder.

01 · Library

Eight agents you can run today

Each agent is a focused specialist with its own prompt, tools and guardrails. Start with the agents that handle the work you hate; trust them in Suggest mode, promote to Auto when you're ready. New agents ship monthly.

Lead Enricher, Inbox Triager, Reply Drafter, Objection Handler
Meeting Booker, Intent Monitor, List Hygiene, Follow-up Pacer
Each agent has a tunable voice + escalation rules
All editable — clone any one and make it yours

Browse all agents

PLATE · 01AI Agentsfig.

Agent roster · workspace6 active

Lead Enricher

Enriched 124 leads · 3m ago

Auto

Inbox Triager

Classified 38 replies · 5m ago

Auto

Reply Drafter

9 drafts awaiting review

Suggest

Objection Handler

Drafted 4 rebuttals · 12m ago

Suggest

Meeting Booker

Held 3 meetings · today

Auto

Intent Monitor

2 hot accounts flagged

Auto

2.1k

Actions/day

24/7

Uptime

11h

Saved/rep/wk

02 · Composer

Chain agents into a workflow

Most outbound work needs more than one agent. A hot account caught by the Monitor needs enriching, then a personalized first touch, then a Booker when the reply lands. DitLead's chain is a real graph: branches, conditions, escalations, all observable.

Drag-and-drop chain editor with branches
Conditional handoffs on confidence, intent, deal value
Per-step memory: every agent reads the chain's context
Replay any historical chain to debug or fork

PLATE · 01AI Agentsfig.

Workflow · inbound qualification5 agents · chained

Intent Monitor

Watches 700M+ DB for hiring + funding signal

1 hot signal

Lead Enricher

Pulls firmographic + tech-stack + emails

Verified

Inbox Triager

Tags inbound by intent: meeting, objection, oos

Routing

Reply Drafter

Writes reply in your voice, references context

Draft ready

Meeting Booker

Proposes 3 windows from your calendar

Held

03 · Trust

Every action is auditable

An agent that does work in the background but won't tell you why is a liability. DitLead logs every input, tool call, reasoning step, output and downstream action — visible per agent, per prospect, per chain. Roll back any single decision in one click.

Full trace per action: inputs, tools, reasoning, outputs
Per-action confidence score + rollback button
Slack + email escalation under your confidence threshold
SOC 2 Type II · single-tenant memory · BYO-LLM available

PLATE · 01AI Agentsfig.

Trace · Inbox Triager · 14:02:111.4s · 2 tool calls

0.0sinputReply from Anna K — 'Sounds interesting, send pricing'

0.1stoolcrm.lookup(prospect=anna-k) → stage: discovery

0.3stoolvoice.profile(workspace=acme) → tone bands loaded

0.5sreasonIntent: pricing-request. Stage: discovery. Confidence: 0.94.

0.7stooltags.write(prospect=anna-k, tag=pricing-asked)

0.9sroute→ Reply Drafter (suggest mode)

1.4sdoneStatus: classified · 1 tag · 1 route

04 · Custom

Build your own — without writing code

Every agent is editable. Open the prompt, swap the tools, tighten the guardrails, change the mode. Or start from blank: a name, a prompt, a tool list, a guardrail, an escalation rule. Custom agents live in your workspace and can be shared.

Visual prompt + tool editor — no code required
30+ first-party tools (CRM, calendar, sequence, search, voice)
Bring your own MCP server for proprietary tools
Publish to the team or keep it private

PLATE · 01AI Agentsfig.

Agent builder · "Pricing-question handler"Draft

PromptWhen a reply asks for pricing, propose 3 windows…

Toolscrm.read · calendar.search · sequence.pause

GuardrailsNever quote discounts. Escalate if deal > $50K.

ModeSuggest (auto after 50 approvals)

MemoryWorkspace · voice profile · deal stage

Escalation→ owner Slack DM if confidence < 0.7

The roster

Every agent your team can run today

Each shipped, edited and run independently. Pick the agents that handle the work you hate.

FEAT / 01

Lead Enricher

Firmographics, tech stack, contact info — pulled and verified per lead.

700M+ DB

FEAT / 02

Inbox Triager

Tags every inbound by intent: meeting, pricing, objection, out-of-scope.

0.4s

FEAT / 03

Reply Drafter

Writes the reply in your voice. References context. Suggest or Auto.

Voice-matched

FEAT / 04

Objection Handler

Drafts rebuttals for the 20 most-common objections, calibrated to deal stage.

20 patterns

FEAT / 05

Meeting Booker

Proposes 3 windows from your calendar, holds them, confirms on accept.

0 friction

FEAT / 06

Intent Monitor

Watches the 700M+ DB for hiring, funding, role-change and tech-stack signals.

Real-time

FEAT / 07

List Hygiene

Cleans dupes, missing fields, role-mismatches, expired contacts.

Per-list

FEAT / 08

Follow-up Pacer

Picks the next-touch timing based on reply latency, channel and prior engagement.

Per-prospect

FEAT / 09

Custom agents

Build your own in the visual editor — prompt, tools, guardrails, escalation.

No code

Benchmarks8 metrics

What agent-powered workspaces hit

Illustrative benchmarks, not measured guarantees — your results depend on ICP, offer, list quality and copy.

ACTIONS / DAY

2.1k

Median across a 6-rep workspace

TIME SAVED

11h/wk

Per rep, median

AUTO-PROMOTE TIME

9 days

Median: Suggest → Auto trust window

INBOX TRIAGE

0.4s

p95 inbound classification latency

BOOKING RATE

+38%

Of qualified replies turn into meetings

ESCALATION RATE

Of agent actions escalate to a human

ROLLBACK RATE

0.6%

Of auto-actions ever get rolled back

CONFIDENCE p50

0.92

Median agent confidence on shipped actions

Built for

From solo founder to RevOps

Different teams need different chains. Same library underneath.

USE / 01

Founders

Run a 5-agent team from day one — enricher, drafter, booker, monitor, pacer.

Pre-built founder voice profile
Slack escalation on every uncertain action
Promote agents week by week as trust grows

USE / 02

Sales teams

Triager + Drafter sit on every rep's inbox. Booker holds meetings. Monitor flags whales.

Per-rep voice profile
Manager view of all agent traces
Per-agent rollout to 1 → many reps

USE / 03

Agencies

Per-client agent profiles. Per-client voice. Per-client guardrails and approval flows.

Per-client memory scopes
White-label trace view
Per-client billing on actions

USE / 04

RevOps

Hygiene + Monitor + Custom 'enrichment auditor' keep the DB clean without manual queries.

Custom-agent SQL + API tools
Daily hygiene reports
Slack incident on data drift

The trade

Specialists you can trust vs. a black box that ships work

Left is what arrives the day you switch. Right is the monolithic AI you stop pretending to trust.

INCLUDEDWhat you get10 items

01Eight specialized agents available the moment you sign in
02Three autonomy modes per agent: Suggest · Auto · Full
03Shared workspace memory — every agent reads the same context
04Drag-and-drop chain editor with conditional branches
05Full per-action trace: inputs, tools, reasoning, outputs
06Confidence score on every output + one-click rollback
07Visual agent builder — prompt, tools, guardrails, no code
08Bring-your-own MCP server for proprietary tools
09Slack + email escalation under your confidence threshold
10Single-tenant memory, BYO-LLM option, SOC 2 Type II

EXCLUDEDWhat you skip10 items

01One monolithic 'AI SDR' you can't inspect
02A black-box that decides whose deal to break
03Hand-built Zapier chains across 9 tools and a Google sheet
04Reps writing the same triage decisions 80 times a day
05Outbound work waiting on whoever opens the inbox first
06Manual lead enrichment with five different scrapers
07Meeting-booking ping-pong over four time zones
08'AI replies' that ship without a voice profile
09No way to roll back a bad auto-action
10Vendor-locked LLMs with no audit trail

vs the alternatives9 rows

The only library of agents you can compose

Most 'AI sales' is one closed product. DitLead is a library of agents you can inspect, chain, audit, and replace.

Capability	AAiSDR	RRegie.ai	111x Alice	Clay	AApollo AI
Specialized agents (not one monolith)
Three autonomy modes per agent	1 mode	1 mode	1 mode	1 mode	2 modes
Visual chain composer		Zapier			Zapier
Per-action trace + rollback				Basic
Build-your-own (no code)
Bring-your-own MCP tools
Shared workspace memory	Per agent	Per agent	Per agent	Per agent	Per agent
BYO-LLM option
SOC 2 Type II

FAQ10 entries

AI agents, explained

Still curious? Our team responds in under an hour.

Talk to sales

A specialized AI teammate built for one job: enriching a lead, triaging an inbox, drafting a reply, handling an objection, booking a meeting, monitoring intent. Each agent has its own prompt, its own tools, its own guardrails — and you can compose several into a single workflow.

A single monolithic 'AI SDR' is opaque and brittle. DitLead splits the work across specialized agents you can inspect, edit, A/B-test, and replace independently. Trust the enrichment agent? Promote it to auto. Don't trust the reply drafter yet? Keep it in suggest-mode. The two coexist.

Yes. Every agent in the library is editable: prompt, tools (CRM read/write, calendar, web search, sequence control), guardrails, escalation rules. Clone an existing agent or start from a blank one. Agents you build live in your workspace; agents you publish can be shared with the team.

You choose, per agent. Three modes: Suggest (draft + you approve), Auto-under-cap (act inside a daily budget without approval), Full auto (no cap, full autonomy). All actions are logged with reasoning so you can audit and roll back any single decision.

Yes. Every agent reads from and writes to the same shared workspace memory — prospect history, campaign goals, voice profile, deal stage, prior touches. The triage agent's tag becomes the reply drafter's signal becomes the booker's calendar window.

30+ first-party tools: CRM read/write, calendar search and hold, sequence control (pause, resume, branch), the 700M+ lead database, voice profiles, web search, the unified inbox, Slack DMs. You can also expose your own internal tools via MCP.

DitLead tracks every agent's approve-rate and edit-rate in Suggest mode. Once an agent is being approved without edits 90% of the time, you get a nudge to promote it to Auto. You can also set custom thresholds.

Yes. DitLead ships with our base model, but Enterprise plans support BYO-LLM (OpenAI, Anthropic, Azure, your own private endpoint). The agent library and trace UI stay the same; only the inference target changes.

Every action ships with a confidence score and a rollback button. One click reverses the action — the CRM update, the calendar hold, the reply sent. The trace stays in the audit log so you can tune the prompt or escalation rule to avoid it next time.

Memory is single-tenant per workspace. You control retention (default 90 days), what gets summarized into long-term memory, and whether prompts and traces are eligible for product analytics. SOC 2 Type II, GDPR-aligned.