PACKWOLF · Waitlist open

Stop doing the work. Lead the pack.

PACKWOLF turns your recurring workflows into a team of AI agents that run on schedule. Map the work once. Assemble the specialists. Flip the switch. You direct the pack; the pack does the work.

Local-first. BYOK from $10/mo at launch. Built by ideius.

Deployment
CloudDesktop
Models
In-houseClaudeOpenAILocalMCP
The pain

Sound familiar? You hired AI to do the work - you're still doing it.

Four pain points every AI power user hits - every Monday, every step, every handoff. PACKWOLF was built to fix all four.

Pain 01

You re-explain the playbook every Monday.

The workflow is in your head. The AI never learns it - you re-teach it every time.

Pain 02

You're the bottleneck.

Every step needs your judgment, every handoff needs your re-briefing. The work can't move without you in the loop.

Pain 03

You hired an assistant, not a team.

One generalist agent. One chat window. One conversation. That's not a team - it's an intern with amnesia.

Pain 04

Nothing connects.

Yesterday's research lives in one tab. This morning's draft lives in another. The chain of work has no chain.

What changes

You stop doing. You start directing.

PACKWOLF turns recurring workflows into a team of AI agents you direct. Three moves get you from doer to pack leader.

Move 01

Map

Draw the workflow once. Steps, branches, decision points - written down. The play stops living in your head.

Move 02

Assemble

Compose the pack. Researcher, drafter, analyst, reviewer - each a specialist with its own memory, tools, and brief. You're hiring, not configuring.

Move 03

Lead

Flip the switch. Heartbeats schedule the work. Memory carries context across handoffs. Approvals catch the moments that matter. You watch the pack work.

How it works

What it takes to run a real AI team.

The team. The playbooks. The runtime. Plus the substrate every agent stands on. Every part is real product, in the same workspace.

Layer 01 · The team

The team you'd hire if headcount were free.

Every agent is a specialist with a curated identity. The org chart organizes them like the team you'd hire if you had infinite headroom. Your mission and values flow into every agent's context, every conversation.

packwolf.app · Team
20 specialists
CLARITY
Chief of Staff
LEX
GC
SENTINEL
CISO
VERA
CSO
DASH
CRO
ROBIN
CMO
NEXUS
CPO
ATLAS
CTO
TRACE
Market Research
GROWTH
Revenue Ops
LAUNCH
App Store
DOCS
Tech Writer
PRISM
QA Engineer
CANVAS
Product Design
NIMBUS
Cloud & DevOps
PIXEL
Frontend Eng
CORTEX
AI Engineer
DART
Flutter Eng
STEEL
Backend Eng
FORGE
Lead Architect
ExecutiveDepartment headsSpecialists
CLARITY
Chief of Staff
LEX
GC
SENTINEL
CISO
VERA
CSO
DASH
CRO
ROBIN
CMO
NEXUS
CPO
ATLAS
CTO
+ 12 specialists across Atlas, Robin, Nexus
ExecutiveDepartment headsSpecialists
Layer 02 · The plan

The playbooks your team runs from.

Twelve plan documents per product, frozen on approval. Four authored specs gate the build manifest. Workflows codify the SOPs - and agents propose new ones from real work. The whole graph breaks down into tasks any specialist can pick up.

packwolf.app · arabic book reader · plans
11/12 frozen
Plans · v1.0 · approved
Strategy & Positioning
Vision, value prop, audience, moats
Frozen v1.0
Apr 20100%
Market Research
TAM/SAM/SOM, competitors, SWOT
Frozen v1.0
Apr 20100%
Business Plan
Revenue, pricing, unit economics
Frozen v1.0
Apr 20100%
Features & Capabilities
Inventory, scope, NFRs
Frozen v1.0
Apr 20100%
Go-to-Market
Launch, growth, channels
Frozen v1.0
Apr 20100%
Marketing Plan
Messaging, content, campaigns
Frozen v1.0
Apr 20100%
Sales Plan
Strategy, segments, pipeline
Frozen v1.0
Apr 20100%
Technical Architecture
Tech stack, API, infra
Frozen v1.0
Apr 20100%
UI/UX Plan
Principles, flows, a11y
Frozen v1.0
Apr 20100%
Decisions Log
Decisions with rationale
Frozen v1.0
Apr 20100%
Metrics & Success
North star, guardrails, KPIs
Frozen v1.0
Apr 20100%
Release Notes
Release history, changelogs
Live draft
Not started0%
Layer 03 · The runtime

The team runs while you sleep.

Scheduled heartbeats decide who runs, when, and at what priority - assess, execute, recover. Agents hand off to each other over A2A; you join by Telegram or web. Every coordination - every generation, every tool call, every delegation - lands in the trace browser as a span.

packwolf.app · scheduling · heartbeats
4 active
Pulse · last 24 hours
sales-prospecting· every 4 hrsP1Next in 0:48
Last run 14:00 UTC·4 prospects logged · 1 hand-off to Sales
docs-audit· every 6 hrsP2Next in 2:48
Last run 12:00 UTC·23 docs reviewed · 3 stale flagged
eval-runs· daily 09:00P1Running now
Last run 09:00 UTC·47 evals passing · 2 regressions
memory-consolidation· daily 03:00P3Next in 7:48
Last run 03:00 UTC·1.2k chunks merged · 18 proposals queued
Priority queue · next runs
1.P1sales-prospectingready in0:48
2.P2docs-auditready in2:48
3.P3memory-consolidationready in7:48
4.P1eval-runsready in14:18
The substrate

What lets the pack run without you.

MemoryCategorized by purpose - colleague, comms, directives, feedback, general, playbook. Consolidated nightly, provenance on every write.
Tools and MCPDozens of built-in tools plus unlimited MCP servers, behind a nine-point safety gate.
Three-tier approvalsAuthor, peer review, sign-off. Versioned, reversible, attributable to a person.
Eval frameworkMulti-skill, multi-model A/B comparison with an LLM grader. Built in.
Costs and budgetsPer-agent spend events. Alert-driven budgets. No bundled surprises.
Workspace sandboxPer-agent filesystem boundaries and scoped tool permissions. Sandboxed by default.
Versus

The questions you'll ask before you commit.

Six tools you might shortlist alongside PACKWOLF. Pick one to compare against - head to head, sources cited.

PACKWOLF vs
Updated May 2026
Daily AI subscriptions
Open-source agent platforms
Workflow + automation
Question
PACKWOLF
ChatGPT Plus
Group 01

Workforce, or another tool?

Whether the agents have identities, talk to each other, work continuously on goals, and propose workflows from what they learn — or whether they're prompts and skill plugins for a single agent doing one task at a time.

Are these named coworkers with their own perspective — or prompts and plugins?Each specialist needs a curated identity (bio, brief, expertise) that evolves as it learns from real work. Coworkers, not configurations.
PACKWOLF
Named teammates with BIO, IDENTITY, brief, HEARTBEAT — markdown identities that evolve through nightly consolidation as the agent learns from real work.
ChatGPT Plus
Custom GPTs are siloed prompts; no persistent identity, no perspective, no evolution.
Do they work in an org chart and talk to each other?A real workforce has managers and reports. Coworkers share context, hand off, escalate. Not all roads lead to you.
PACKWOLF
Org chart with managers and reports; agent-to-agent messaging over A2A. You join threads by Telegram or web when the pack needs you.
ChatGPT Plus
One ChatGPT per conversation; no peer-to-peer.
Will they work on a goal for weeks — or just on a single prompt?An employee picks up a project Monday and keeps moving on it until Friday. A tool runs when triggered and stops.
PACKWOLF
Heartbeats with priority scoring across the whole pack; agents work goals that take weeks, surfacing decisions when they need you.
ChatGPT Plus
Scheduled Tasks in beta — 10 active max, single-assistant, task-shaped not goal-shaped.
OpenAI · Tasks ↗
Do new workflows emerge as agents work — or do you have to pre-author everything?A real team learns by doing. They notice patterns, propose playbooks, refine them. You approve the ones worth keeping.
PACKWOLF
Agents propose new workflows from learned patterns; you approve through three-tier sign-off. The pack discovers playbooks you didn't write.
ChatGPT Plus
Memory remembers preferences; no workflow proposals.
Will the pack remember what we decided last quarter?Memory categorized by purpose so the right specialist surfaces the right context at the right turn — not a single bucket of everything.
PACKWOLF
Four-layer memory (working, episodic, durable, transcript) categorized by purpose; shared across web, desktop, and Telegram with provenance per chunk.
ChatGPT Plus
Cross-chat memory with a Sources provenance panel; account-scoped, single bucket.
OpenAI · Memory ↗
When something goes wrong, can I see exactly what every agent did?Per-span trace, prompt versioning, replay — not just chat history or run logs.
PACKWOLF
Per-span flame graph, prompt versioning, replay, full audit log on every tool call across every agent.
ChatGPT Plus
Chat history only.
Group 02

On your terms?

Where it runs, whose models it uses, whether you can switch deployments, and how hard it is to get started.

Does my workspace, memory, and trace data live on my machine?Local-first means more than 'a desktop app exists.' Means workspace + memory + traces stay on your hardware.
PACKWOLF
Desktop app on every tier; workspace, memory, and traces all stay local by default.
ChatGPT Plus
Cloud only — chats, memory, files all live with OpenAI.
Can I use my own model keys without markup?Bring your own keys across providers — your provider bills go to the provider, not to us with a margin.
PACKWOLF
BYOK across Claude, OpenAI, local LLMs, MCP servers — $10/mo entry tier, no markup.
ChatGPT Plus
OpenAI models only.
Can I use cloud and desktop without giving up control of my data?BYOK keeps the work local. Basic runs the whole operation locally with cloud visibility, approvals, included usage, and BYOK overflow. Pro and Max let approved agents run on PACKWOLF Cloud when selected work should keep moving without the desktop.
PACKWOLF
Cloud and Desktop ship as one product. BYOK stays local, Basic runs locally with a cloud command center and BYOK overflow, and Pro/Max let approved agents run on PACKWOLF Cloud.
ChatGPT Plus
Cloud only.
Can a non-engineer get this running today?Five-minute install for prosumers without devops experience — or self-host setup with Docker, Node, and credentials?
PACKWOLF
Download and run. Five-minute setup, no devops required. Same for Cloud (browser signup).
ChatGPT Plus
Cloud signup — instant.
Wins this rowPartial - does some of thisHas it, in their own way-Not in product

Comparison reflects publicly documented capabilities as of May 2026. ChatGPT, Claude, Cursor, Lindy, n8n, and Zapier are trademarks of their respective owners.

Common questions

What you'll want to know before you join.

We're shipping the waitlist preview to early cohorts now, with public launch later this year. Waitlist members get early access and locked-in pricing - whatever tier you pick today is the tier you pay at launch.

What you get on day one.
Local-first by default
Your data stays on your machine. Cloud coordinates unless you choose continuity.
BYOK
Your keys, no markup. Bring Claude, OpenAI, local models, or your own MCP servers.
Sandboxed per agent
Per-agent filesystem boundaries and scoped tool permissions.
No lock-in
Export everything, anytime. Your workflows, your memory, your agents.
Built by ideius
A real team with a real track record building AI for real businesses.
Read the full security posture →
Pricing

Simple pricing. Same runtime.

Start local with your own keys. Add our model and command center. Scale when approved agents need to keep running on PACKWOLF Cloud.

Local app
BYOK

Your keys. Your machine. No included usage.

  • Full local runtime
  • Account + license
  • Your provider bills
$10Per month at launch
PopularCommand center
Basic

Full local operation, plus our model and sandbox credits.

  • 1x included usage
  • 10 E2B hours
  • BYOK overflow
$30Per month at launch
Pro / Max
Scale

Approved agents can run on PACKWOLF Cloud.

  • Pro: 5x Basic usage
  • Max: 20x Basic usage
  • Cloud-run agents
$100+Per month at launch
See full pricingJoin the waitlist

Stop doing the work. Lead the pack.

Join the waitlist. Lock in your tier. We'll open the gate as soon as we're ready.

Join the waitlist