Abe Stack

Flagship offering

Agentic AI & Intelligent Automation — software that actually thinks and acts.

Custom AI agents that handle real workflows — not just chatbots. Tool-calling agents on your Firebase, Gmail, and Sheets stack, with humans approving before anything goes live.

Get a free AI opportunity audit Try live demo
Genkit / ADK Tool-calling agents Firebase Multi-LLM failover

What we build

Agentic systems for teams that outgrow Zapier — and chatbot widgets.

Every agent is scoped to your ops, wired to your tools, and deployed on infrastructure you own.

Custom agents (Genkit / ADK)

Purpose-built agents on Google's Genkit and Agent Development Kit patterns — orchestrated prompts, evals, and production guardrails, not generic GPT wrappers.

Tool-calling agents

Agents that call Gmail, Sheets, Firestore, social APIs, and webhooks — read data, take action, and log every step for auditability.

Multi-agent systems

Specialist agents that hand off to each other — research → draft → review → publish — coordinated by queue workers, not a single overloaded prompt.

RAG + automation

Retrieval-augmented agents over your docs, SOPs, and Firestore data — combined with scheduled jobs and event triggers so answers turn into actions.

Process

How it works

Same rigor we bring to every Abe Stack project — adapted for agentic systems.

01

Discover & map agents

Identify workflows worth automating, define agent boundaries, tools, and human approval gates before any code runs.

02

Design orchestration

Prompt pipelines, tool schemas, multi-agent handoffs, and failover routing — documented so you know exactly what ships.

03

Build & review gates

Implement on Firebase with daily Loom updates. Every external action passes through a human review step in staging first.

04

Ship & monitor

Production deploy, queue monitoring, error alerts, and handoff docs — plus optional care plans for prompt tuning and new tools.

Use cases

Agents we've built — and patterns ready for your ops.

Inquiry Triage Pipeline

100% inquiries queued with AI category + urgency summary before human reply. Case study →

Publishing Agent

85% less rewriting · 1,200+ publish actions. Shipped on Prayer City Social.

Why Abe Stack

Benefits & differentiators

  • ✓Production-ready — not a demo that breaks on week two. Queue workers, retries, monitoring, and handoff docs included.
  • ✓Your stack, connected — Firebase, Gmail, Sheets, OAuth social APIs, and webhooks — agents plug in, not replace.
  • ✓Owned by you — source code in your GitHub repo. No per-seat AI SaaS, no vendor lock-in on prompts or data.
  • ✓Human-in-the-loop — agents draft and recommend; your team approves before emails send or posts publish.
  • ✓Senior technical leadership — dedicated architect on your agent systems, with documented runbooks, staging environments, and escalation paths — without agency markup.

Transparent pricing

AI Agent Add-on

Scoped after discovery — fixed quote, no hourly surprises. Add to a Platform project or standalone workflow build.

$4,000 – $12,000

  • ✓ 1–2 production agent workflows
  • ✓ Tool-calling + queue workers
  • ✓ Multi-LLM routing with failover
  • ✓ Human review dashboard
  • ✓ Handoff docs + 60-day support window

Platform projects (auth + dashboards): $8K–$15K+ · Full pricing →

Get a free AI opportunity audit

Try it yourself

Volunteer Coordinator Agent — interactive demo.

Describe a coordination goal and watch the agent stream reasoning steps — roster load, shift matching, and human-in-the-loop approval queue. Fictional sample data only; production uses Genkit + Firebase.

Live public demo. Responses adapt to keywords in your goal (try Spanish greeters, kids ministry, parking, tech booth, or first aid). Fictional sample data only — no real sends; all actions queue for coordinator approval.

Submit a goal to see reasoning steps stream here.

Reasoning trace and tool calls appear after you run the agent.

How this pipeline works → · Volunteer Matching case study →

Agents in production

Screenshots from shipped agent workflows

Real agent consoles from Prayer City ops — digest intelligence dry runs, volunteer matching, and multi-platform publishing.

Daily Digest Intelligence Agent dry run — pipeline, metrics, and AI analysis summary
Daily Digest Intelligence Agent · Genkit + Firebase · dry run (no emails sent)
Agent-suggested actions with priority flags and human approval gates
Suggested actions — FLAG_MISMATCH, NOTIFY_ORGANIZER, ASSIGN_TENT_REMINDER with approval gates
Volunteer signup data feeding the matching agent roster tool
Volunteer Matching Agent · 70% less manual matching
Publishing agent dashboard with queued variants awaiting approval
Prayer City Social · 1,200+ publish actions
In-browser video editor integrated with the publishing agent pipeline
Agentic Publishing Layer · 85% less rewriting

See it in action

Agent dry run — Daily Digest Intelligence.

Watch a real dry run: the agent reads the volunteer roster, runs Gemini analysis, surfaces mismatches and risks, and queues suggested actions — dryRun: true, no emails sent until a coordinator approves.

Prayer City automation case study →

Agent dry-run walkthrough

Loading Loom… or try the live demo above ↑

Ready to build agents that actually work?

Start with a free AI opportunity audit or book a 30-minute discovery call. Fixed quote within 48 hours if we're a fit.

Get a free AI opportunity audit Send a project inquiry