Pancake

Overview

Pancake is an autonomous AI agent system that lives inside Slack. The pitch is to give a founder or small company an "AI org chart" — squads of agents organized by department (growth, engineering, operations) that run real tasks 24/7, sync context from meeting notes and Slack discussions, and escalate to a human only when an action crosses a spend, scope, or trust threshold.

Each agent is markdown-configured: behavior is defined in .md files rather than a visual builder, which fits teams that want their AI agent definitions version-controlled and editable like code. Agents are context-aware (reading Notion, docs, and meeting notes), always-on, sandboxed with tool/data restrictions, and backed by audit logging with replay and rollback.

Pancake targets founders and startups that want operational leverage without hiring — the framing is "your co-founder," with agent squads that can write copy, ship pull requests, screen recruits, handle invoicing, and post daily digests. Both the official site and Product Hunt position Pancake as an OpenClaw-based Slack cofounder layer; Pancake's own blog describes it as the assembled company layer built on top of Hermes and OpenClaw.

Key Features

• Agent squads by department — Pre-organized squads for Growth (copywriter, ad manager, social, email), Engineering (full-stack, DevOps, performance monitor, QA), and Operations (scheduling, recruiting screener, invoicing, support), so the org chart maps to real functions rather than a generic agent.
• Markdown-configured agents — Agent behavior is defined in .md files, making definitions version-controllable and editable like code rather than locked in a visual builder.
• Lives in Slack — Agents operate inside Slack where work is already discussed, with one-tap approvals on phone or Slack for actions exceeding spend, scope, or trust thresholds.
• Context-aware and always-on — Agents read Notion, docs, and meeting notes for context and run 24/7, pairing naturally with coding workflows like Claude Code for the engineering squad's pull-request work.
• Approval gates for risky actions — Irreversible or costly actions (spend, production changes) require explicit human approval, so autonomy doesn't mean unsupervised.
• Audit logging with replay/rollback — Every agent action is logged with replay and rollback, and agents are sandboxed with tool/data restrictions — important when a fleet of agents acts on company systems.

How It Compares

• vs. Lindy / Cassidy / general AI agent builders — Those give a canvas to build individual automations. Pancake ships a pre-organized org chart of department squads and lives in Slack, trading some build flexibility for faster time-to-an-operating-team.
• vs. Manus / autonomous task agents — Manus excels at completing a single complex task end-to-end. Pancake is built for ongoing, parallel operations across functions with persistent context, closer to "standing team" than "one-shot task."
• vs. Slack workflow bots (Standuply, Workflow Builder) — Those are scripted automations. Pancake's agents are LLM-driven, context-aware, and markdown-configured, with audit/rollback — a different abstraction than trigger-action bots.
• vs. coding-agent-only tools (Claude Code, Codex) — Those focus on the engineering loop. Pancake spans growth, ops, and engineering, using the coding agent as one squad among several.

The honest read: Pancake is best for founders who want a standing, multi-function agent team operating in Slack. It's a worse fit for teams that need one deep capability (just coding, just sales) or that don't run on Slack.

Pricing & Plans

Pancake uses a token-pass-through model — "we charge what the labs charge," with margin coming from volume discounts. There's a 3-day trial with $100 in credits (no credit card required), then paid plans.

Token costs are pass-through to the underlying model labs; the monthly fee covers the platform/infrastructure rather than marked-up tokens. Tokens reset monthly, token-pack changes are prorated, and cancellation is unrestricted. Because real cost tracks token consumption, benchmark a representative workload before committing.

Best For

• Founders and tiny startups wanting an operating team across growth, engineering, and ops without hiring
• Slack-native teams comfortable approving agent actions from their phone
• Technical founders who want agent definitions as version-controlled markdown rather than a visual builder
• Operators who need AI productivity leverage across multiple functions, not just one deep capability
• Teams comfortable with usage-based token costs in exchange for no token markup

FAQ