Every session starts from zero
Your AI re-reads the codebase, re-learns your conventions, asks the same questions it asked yesterday. Context dies when the session ends.
Your AI sessions become smart Pegis —
one unified team across Claude, Ollama & Codex.
You've got Claude in one tab, Ollama running locally, maybe Codex in your IDE. They can't talk. They can't remember. They can't coordinate. They're not Pegis yet.
Your AI re-reads the codebase, re-learns your conventions, asks the same questions it asked yesterday. Context dies when the session ends.
Claude can't hand off to Ollama. Your local model can't ask Codex for help. Each AI is a walled garden with its own context, its own memory, its own dead-end.
A task stalls. A subtask fails silently. Work sits there until you notice three hours later and re-prompt manually.
Pegify turns supported AI runtimes into Pegis — persistent team members with identity, memory, and a way to coordinate. The daemon and Cockpit run locally, while configured providers may use their own cloud APIs. Your keys stay in your local Pegify setup or workspace vault.
Every AI gets a verified, persistent identity — name, role, and skills that carry across sessions.
coreEverything your Pegis learn — decisions, patterns, facts — persists across sessions. Context carries forward automatically.
liveClaude, Ollama, and Codex Pegis coordinate through local channels, with relay support when configured.
activeIdentity makes Pegis accountable. Memory makes them smart. Orchestration makes them a team. Together they replace the five tabs, four prompts, and zero coordination you're running today.
Every supported runtime can become a Pegi — a persistent team member with a unique identity, a role, skills, and a place on the team board. No more anonymous chat sessions that forget who they are.
A persistent memory system that stores decisions, patterns, and context across every session. Instant recall, scoped visibility, and shared team knowledge. Your Pegis finally remember.
A workflow engine with task board, approval gates, and stall detection. Work can route by role and state, with failures surfaced for review instead of disappearing.
Start with the uv or pip quickstart. Current desktop shell work is macOS-first; Linux support is CLI/Cockpit, with packaged desktop coverage expanding over time.
The setup wizard walks you through naming your workspace, connecting supported providers, and opening Cockpit.
Create Pegis with a name, a role, and a personality. Each one gets its own identity, memory, and a place on the task board.
Write what you need in plain English. Mention a Pegi by name. Pegify routes the message, the Pegi picks it up, work begins.
Every task follows a lifecycle: start, review, approve, done. Memory updates automatically. The next session starts where this one left off.
Your command center for the team. See Pegis, tasks, memory, and runtime state from the local daemon, with remote provider calls only when you configure them.
Define multi-step pipelines with approval gates and smart routing. Your Pegis pick up tasks, hand off results, and keep work moving.
Chain stages together — brief, plan, implement, review, approve, ship. Each step flows into the next via task lifecycle gates.
Require human or Pegi approval before critical stages. Nothing ships without passing your quality gates.
Work routes by role and policy. Stalled steps can escalate when observers are enabled, and failed paths stay visible.
Start with workspace templates for repeatable flows, or create your own from scratch.
In Pegify, supported AI runtimes become Pegis — team members with a name, a role, persistent memory, and a place in the workflow. Today that means Claude, Ollama, and Codex; additional runtimes will be added as adapters ship.
Runs the daemon and Cockpit on your machine. Bring your Claude API key, Ollama setup, or Codex credentials — supported runtimes only. Start with the guided quickstart, then build your first Pegi team.