What Is Fleece AI? Agent Platform Explained
By Loïc Jané · Founder, Fleece AI
What Is Fleece AI?
At a Glance: Fleece AI is a delegative AI workspace powered by autonomous, self-improving AI agents that automate multi-step workflows across 3,000+ apps — no code required. Agents communicate, delegate tasks to each other, and optimize their own prompts over time. Updated September 2026.
Fleece AI is an autonomous AI agent platform built for intelligent workflow automation — explained in full below. Gartner predicts 33% of enterprise software will include agentic AI by 2028 — Fleece AI is already there. You describe a task in natural language, and Fleece AI deploys autonomous AI agents that execute multi-step workflows across 3,000+ app integrations — 24/7, on a schedule, with no code required.
Fleece AI is not a chatbot. It is a delegative AI workspace where you assign goals to autonomous AI agents and they deliver results without ongoing supervision. Think of it as hiring a digital team that connects to your Gmail, Slack, Notion, Stripe, Salesforce, and thousands of other tools — then does the work for you, learns from each execution, and gets better over time.
Why Is It Called Fleece AI?
The name Fleece AI evokes the idea of the Golden Fleece — a prized, sought-after solution. In Greek mythology, the Golden Fleece represented the ultimate quest. In the AI space, Fleece AI represents the goal every business is chasing: autonomous execution of tedious work by intelligent agents that continuously improve.
The official website is fleeceai.app.
What Makes Fleece AI Different: Self-Improving Autonomous Agents
Most AI automation tools execute static workflows. Fleece AI agents are fundamentally different — they are autonomous, self-improving, and collaborative.
Self-Improving Agents
Fleece AI agents optimize their own performance over time. Manager agents can automatically update their sub-agents' system prompts based on execution results — a capability called prompt auto-improvement. Every modification is logged with full history, so you can review changes and rollback if needed. This means your AI agents get smarter with every task they complete.
Inter-Agent Communication & Hierarchical Delegation
Fleece AI features a multi-agent system with hierarchical delegation. You can organize agents into teams with manager-subordinate relationships:
- Manager agents delegate tasks to specialized sub-agents and receive structured reports back
- Sub-agents report results to their parent agent automatically
- Broadcast messaging lets a manager coordinate an entire team at once
- Delegation depth limit bounds delegation chains — the system enforces a maximum delegation depth of 3 levels
This is not just chat between bots. It is a structured orchestration layer where each agent has a defined role, clear responsibilities, and accountability.
Heartbeat Scheduling (Autonomous Periodic Execution)
Fleece AI agents can be configured with heartbeat schedules — autonomous wake-up intervals from every 5 minutes to every 24 hours. An agent with a heartbeat wakes up on its own, performs its assigned task, and goes back to sleep. No triggers, no cron jobs to configure manually, no human in the loop.
This is ideal for monitoring tasks, periodic reporting, data synchronization, and proactive alerting.
Custom Skills & Knowledge Files
Every Fleece AI agent can be equipped with custom skills (user-created capabilities with specific instructions) and knowledge files (documents injected directly into the agent's context). With 30+ built-in skills plus unlimited custom skills, each agent becomes a specialist in its domain.
What Does Fleece AI Do?
Fleece AI lets you automate any recurring business workflow by describing it in plain English (or French — the platform is fully bilingual). Here is what that looks like in practice:
1. You describe the task "Every Monday at 9 AM, summarize my unread Gmail messages and post the summary to my Slack #updates channel."
2. Fleece AI connects your tools The platform uses managed OAuth to connect to Gmail and Slack — no API keys, no developer setup. Over 3,000 app integrations are available via Pipedream (20+ essential apps on Starter, 1,000+ on Pro, the full catalog on Business).
3. Fleece AI executes autonomously An autonomous AI agent runs every Monday at 9 AM, reads your inbox, generates a summary, and posts it to Slack. You never touch it again.
4. Your agents learn and improve Over time, manager agents can refine how sub-agents work — adjusting prompts, reallocating tasks, and optimizing the entire workflow chain automatically.
Core Capabilities
| Feature | Details |
|---|---|
| Integrations | 3,000+ apps with managed OAuth (Gmail, Slack, Notion, Salesforce, Stripe, HubSpot, etc.), tiered by plan |
| Scheduling | Cron-based recurring flows + agent heartbeat auto-scheduling |
| AI Models | GLM 5.3 Flash (default), GPT-5.2, GPT-5.6 Luna and Mistral Medium 3.5 (Starter); GPT-5.6 Terra (Pro); GPT-5.6 Sol and Claude Opus 5.5, 4.8 and 4.7 (Business); Claude Fable 5 (Enterprise) |
| Multi-Agent System | Hierarchical delegation, inter-agent communication, self-improving prompts |
| Custom Skills | 30+ built-in + unlimited user-created skills per agent |
| Knowledge Files | Attach .md/.txt documents directly to agent context |
| Multi-Step Workflows | Multi-step executions with dynamic tool selection |
| Setup Time | Under 60 seconds, no code required |
For a complete breakdown of everything the platform and its agents can do, see the full Fleece AI capabilities overview.
Try Fleece AI for 4 days — Deploy your first autonomous agent in under 60 seconds.
Fleece AI vs. Other Things Named "Fleece"
If you searched for "Fleece" and ended up here, you might have been looking for something else. Here is how Fleece AI compares to other uses of the name.
Fleece (Fabric / Clothing)
The most common meaning of "fleece" is a type of soft fabric or outerwear made from polyester or wool. Fleece jackets, fleece blankets, and fleece pullovers are popular in outdoor clothing.
Fleece AI has nothing to do with textiles. It is a software platform for autonomous AI agent orchestration. The name is inspired by the mythological Golden Fleece, not the fabric.
keldenl/fleece (GitHub — VS Code Extension)
There is an open-source project called fleece on GitHub by keldenl. It is a VS Code extension for LLM-powered code generation that uses local language models (via llama.cpp) for inline code suggestions.
How Fleece AI differs: Fleece AI is a full autonomous agent platform, not a code editor extension. While keldenl/fleece focuses on developer productivity within VS Code, Fleece AI automates cross-app business workflows with self-improving agents for technical and non-technical users alike.
couchbase/fleece (Binary Data Format)
Couchbase maintains a library called Fleece, which is a high-performance binary encoding format for semi-structured data. It is used internally by Couchbase Lite for efficient JSON-like data serialization.
How Fleece AI differs: This is a low-level data format library for database internals. Fleece AI is an end-user platform for automating business tasks with autonomous AI agents.
fsprojects/Fleece (F# JSON Library)
In the F# programming ecosystem, Fleece is a JSON mapping library for serializing and deserializing data structures.
How Fleece AI differs: This is a developer library for a specific programming language. Fleece AI is a no-code autonomous agent platform accessible to anyone.
rackerlabs/fleece (Serverless Framework)
Rackspace published a Python framework called Fleece for building serverless applications on AWS Lambda.
How Fleece AI differs: While both involve automation, rackerlabs/fleece is a developer framework for writing Lambda functions. Fleece AI provides autonomous, self-improving agents that require no code at all.
How Fleece AI Works Under the Hood
Fleece AI uses a delegative AI architecture combined with a multi-agent orchestration system. Here is what that means:
-
Natural Language Understanding: When you describe a task, Fleece AI parses your intent and identifies the apps, data sources, and actions involved.
-
Tool Discovery via MCP: Fleece AI agents discover available integrations dynamically through the Model Context Protocol (MCP). They do not rely on pre-built templates — the AI selects the right tools for each task.
-
Managed Authentication: Every app connection uses OAuth managed by Fleece AI. You click "Connect," authorize the app, and Fleece AI handles token storage and refresh.
-
Autonomous Execution: Once deployed, a Fleece AI agent runs independently — on a cron schedule or via heartbeat auto-scheduling. It uses the model you pick for that agent — GLM 5.3 Flash by default, up to Claude Opus 5.5 on Business — to reason through each step.
-
Inter-Agent Orchestration: For complex workflows, multiple agents collaborate through hierarchical delegation. A manager agent breaks down the goal, delegates sub-tasks, collects reports, and synthesizes the final result.
-
Self-Improvement Loop: Manager agents analyze execution results and can automatically refine sub-agent prompts — improving accuracy, reducing errors, and adapting to new patterns over time. All changes are logged for transparency.
-
Error Handling: If a step fails, Fleece AI retries with exponential backoff. If the failure persists, it notifies you with context about what went wrong.
Import Your Existing AI Agents
Fleece AI used to offer a one-click import of OpenAI Assistants. OpenAI sunset the Assistants API on August 26, 2026, so there is nothing left to import from. To move an existing assistant or custom GPT, paste its instructions into a new Fleece AI agent, attach its reference documents as knowledge files, and connect the apps it needs.
Who Uses Fleece AI?
Fleece AI is designed for anyone who has recurring tasks that span multiple business tools:
- Founders and solopreneurs who want to automate without hiring — let self-improving agents handle operations
- Operations teams managing cross-app workflows (CRM updates, report generation, data syncing) with multi-agent delegation
- Marketing teams automating content distribution, social monitoring, and competitor tracking with scheduled heartbeat agents
- Sales teams automating lead enrichment, follow-ups, and pipeline reporting with specialized agent hierarchies
- Support teams triaging tickets, generating summaries, and escalating urgent issues through inter-agent communication
- Enterprise teams building custom agent hierarchies with manager-subordinate structures for complex workflows
No coding is required. If you can describe the task, Fleece AI can automate it — and improve at it over time.
Getting Started with Fleece AI
Setting up Fleece AI takes under 60 seconds:
- Start your 4-day trial at fleeceai.app
- Create your first autonomous AI agent
- Connect your apps — Gmail, Slack, Notion, Stripe, and 3,000+ more
- Describe your workflow in natural language or set up a heartbeat schedule
- Done — your Fleece AI agent handles it from here, and gets smarter with every execution
Frequently Asked Questions
Is Fleece AI free?
Fleece AI offers a 4-day free trial (card required, no charge until it ends) that includes 1,000 credits, 5 agents and 10 flows. After the trial, plans start at €49/month (or €39/month with yearly billing — 20% savings) for Starter with 2,000 credits per month; Pro is €99/month and Business €199/month.
What are autonomous AI agents?
Autonomous AI agents are AI systems that can plan, execute, and adapt to tasks independently — without step-by-step human guidance. Fleece AI's agents go further with self-improving capabilities: they analyze their own performance and optimize their approach over time.
What AI models does Fleece AI use?
GLM 5.3 Flash is the default model. Starter also includes GPT-5.2, GPT-5.6 Luna, Mistral Medium 3.5, Qwen 3.8 Max, Kimi K3 and DeepSeek V4.1 Flash; Pro adds GPT-5.6 Terra; Business adds GPT-5.6 Sol and Claude Opus 5.5, 4.8 and 4.7; Enterprise adds Claude Fable 5. Model usage is included in plan credits, and you can choose the model per agent.
What is inter-agent communication?
Inter-agent communication is Fleece AI's system for agents to delegate tasks, share results, and coordinate work. Manager agents can assign sub-tasks to specialized agents, receive reports, and even optimize their sub-agents' prompts automatically. This enables complex, multi-step workflows that no single agent could handle alone.
Can I import my OpenAI Assistants?
Not any more. OpenAI shut down the Assistants API on August 26, 2026, which removed the source the one-click import relied on. You can recreate an assistant in a few minutes by pasting its instructions into a new Fleece AI agent and attaching its documents as knowledge files.
Does Fleece AI replace Zapier or Make?
Fleece AI is AI-native, meaning you describe tasks in natural language instead of building trigger-action workflows manually. Unlike Zapier or Make, Fleece AI agents are autonomous and self-improving — they adapt and optimize without manual intervention. For detailed comparisons, see Fleece AI vs Zapier and Fleece AI vs Make.com.
Is Fleece AI safe to use with my business data?
Yes. Data is encrypted in transit and at rest, SOC 2 Type II certification is in progress, and Fleece AI provides a DPA and Standard Contractual Clauses for GDPR. Your data is never used to train AI models. Inter-agent communication is scoped per user — agents cannot access data from other accounts.
Related Articles
- What Is Delegative AI? — the paradigm behind Fleece AI
- AI Workflow Automation Guide — complete guide for 2026
- Best AI Agent Platforms — full market comparison
Try Fleece AI's 4-day trial — deploy your first autonomous, self-improving AI agent in under 60 seconds at fleeceai.app.