Agent sessions — message routing, memory scoping, topic forking, multi-agent coordination. Routes messages to folders, isolates context per team, composes tools via MCP.
Multi-tenant agent context orchestration primitives: topic-suffix routing (target=corp/eng#alerts), folder-per-team isolation (container boundaries), sibling observation (context sharing without secret leakage), session forking (cp -r .claude/).
Multi-agent teams with spatial canvas UI. Visual workspace where agents have status, communicate directly, and coordinate across repos. Desktop-based agent coordination.
Spatial canvas for agent visualization and coordination. See your whole team at a glance: who's active, who's waiting, who's idle. Agent-to-agent communication (DMs, group channels, structured handoffs) with visual feedback.
Distributed workflows with durable execution. Guarantees workflows complete despite infrastructure failures, persists state at each step, resumes from last checkpoint on different worker.
Durable execution with replay guarantees. Replaces manual retry logic + state tracking DB + cron scheduler. Event sourcing for crash recovery, structured retry policies, automatic timeouts.
Data pipelines — tables, datasets, ML models, reports. Asset-first orchestration where data assets are first-class citizens with clear lineage and dependencies.
Asset-first orchestration (not task-first). Unified control plane for data platforms with integrated observability. Multi-tenancy via code locations (subprocess + virtualenv isolation).
Python workflows with flexible execution. Task-based model (not asset-first). Runs your Python functions as-is in your infrastructure — add @flow decorator and done.
Python-first orchestration with zero marginal cost (charges by users/workspaces, not task executions). Dynamic, event-driven workflows. Operational resilience without opinionated data model.
Batch data processing — ETL/ELT, data integration, ML training workflows. Battle-tested at scale with well-defined start and end. DAG-based orchestration.
Battle-tested batch orchestration at massive scale. Largest ecosystem for data integrations. Regulatory compliance in financial services (Dodd-Frank, GDPR reporting).
Integration workflows — app-to-app automation, API calls, webhooks, form submissions. Node-based canvas with 70+ AI-specific nodes (LLMs, embeddings, vector DBs, OCR).
Self-hosted AI-native automation platform. Charges per workflow execution (not per task), dramatically cheaper for complex multi-step automations. LangChain integration with 70 AI nodes.
Personal productivity automation — 7,000+ integrations, no-code. Connects applications at surface level (form submission → email, Slack → Notion). Charges per task.
Largest integration breadth (7,000+ apps). No-code accessibility for non-technical teams. AI Agents (autonomous AI teammates) and Copilot (AI-powered builder) in 2026.
UI-triggered automation for internal tools. Extension of Retool's internal tool builder, not standalone automation engine. Background tasks tied to user interfaces.
Tight integration between UI components and workflow engine. Natural extension for teams already standardized on Retool for admin panels, ops dashboards, approval flows.
Agent memory architecture — tiered memory (core/archival/recall). OS-inspired model: core memory (always in-context, like RAM), archival memory (vector store queried explicitly), recall memory (conversation history).
Tiered memory architecture for stateful agents. Agents learn and self-improve over time. #1 model-agnostic OSS harness on TerminalBench. Conversations API for shared memory across parallel experiences (Jan 2026).
Organizational memory aggregation — cross-tool context syncing (Fathom meetings, Slack, Granola notes, Claude/Codex outputs). Background organization by project/person/decision. MCP-native distribution.
Shared context layer (not internal memory architecture). Multi-source capture, background organization, shared recall across agents + humans. Memory portable across model/client boundaries via MCP.
| Platform | Domain | Orthogonal Contribution | What It Replaces |
|---|---|---|---|
| Temporal | Distributed execution | Durable workflows with replay guarantees | Queues + state DB + cron + manual retry logic |
| Dagster | Data pipelines | Asset-first orchestration with lineage | Task-first schedulers + scattered metadata |
| Prefect | Python workflows | Flexible execution in your infra | Airflow complexity + vendor lock-in |
| Airflow | Batch data processing | Battle-tested ETL at scale | Custom cron jobs + dependency management |
| n8n | Integration automation | Self-hosted AI-native workflows | Zapier SaaS + privacy concerns |
| Zapier | Personal automation | 7,000+ integrations, no-code | Manual data entry + copy-paste between apps |
| Retool Workflows | UI-triggered automation | Tight UI integration for internal tools | N/A (only for Retool users) |
| arizuko | Agent context | Multi-tenant routing + isolation | Manual agent deployment + context leakage |
| Pentagon | Multi-agent teams | Spatial canvas with agent communication | Terminal windows + manual coordination |
| Letta | Agent memory | Tiered memory (core/archival/recall) | Flat vector stores + no learning over time |
| Memory Store | Organizational memory | Cross-tool aggregation + MCP distribution | Fragmented docs + siloed tool contexts |
Shutdown date: December 31, 2025
Architecture: FastAPI + Temporal + PostgreSQL + TimescaleDB + TypeSpec
Promise: "Serverless platform for AI workflows and agents"
What they built:
The fatal flaw: Julep wrapped Temporal + PostgreSQL with no novel primitive. When asked "what does Julep do that Temporal doesn't?", the answer was ceremony:
Oracle test failure: "Why use Julep vs Temporal + Supabase?" → No answer beyond "we wrote docs."
Lesson: Ceremony is not orthogonality. Users could build Julep's functionality with Temporal + Supabase in 3 days. Hosted service couldn't justify pricing when self-hosting the underlying primitives was trivial.
Pivot: Team launched memory.store (YC P26) — cross-tool aggregation with MCP distribution. Actually orthogonal.
arizuko doesn't compete with agent frameworks — it orchestrates them. Same relationship as Dagster (control plane) + dbt (execution engine) in data pipelines.
What you bring:
What arizuko provides:
key=glob patterns)Architecture stack:
Messaging Platforms (Slack, Telegram, Discord, WhatsApp)
↓
[arizuko routing + isolation]
↓
Agent Frameworks (Claude Code, LangGraph, Temporal, Julep)
↓
Execution (LLM calls, tool use)
Analogy: Kubernetes orchestrates containers. arizuko orchestrates agent sessions.
| Platform | Control Plane | Execution | Metadata Storage | Data Residency |
|---|---|---|---|---|
| Dagster+ | Managed SaaS | Customer-hosted agents | Dagster-hosted DB | Metadata leaves env |
| Prefect Cloud | Managed SaaS | Customer infra | Prefect-hosted | Metadata leaves env |
| Zapier | Managed SaaS | Zapier-hosted | Zapier-hosted | All data leaves env |
| n8n Cloud | Managed SaaS | n8n-hosted | n8n-hosted | All data leaves env |
| Temporal Cloud | Managed SaaS | Customer workers | Temporal-hosted | History leaves env |
| Airflow (Astronomer) | Managed SaaS | Customer-hosted | Astronomer-hosted | DAG defs leave env |
| Retool Cloud | Managed SaaS | Retool-hosted | Retool-hosted | All data leaves env |
| arizuko | Self-hosted | Self-hosted containers | Local SQLite | Nothing leaves box |
| Letta | Self-hosted | Local agents | Local vector DB | Nothing leaves box |
| Pentagon | Local desktop | Local agents | Local filesystem | Nothing leaves box |
Existing Research:
Web Research (2026-05-20):