A GUIDE TO THE AGENT CATALOGUE
The agent atlas
Explore how AI agents work and find out which ones fit your task, technology and budget.
Find agents for your task
For a one-off search, an agent you start yourself is often enough. A daily briefing needs an agent that starts on a schedule. Select a task first. The map then shows how much setup the matching agents need. Open a logo to see the reasoning and requirements in its profile.
Autopilot / Scout
Lindy
Zapier Agents
n8n AI Agents
Copilot Studio
Agentforce
Devin
Gemini agent
Webex AI Agent / Dialog
Augment Cosmos
Dify
Dust
Sierra
Decagon
Maritime
Gumloop
Glean Agents
Claude Managed Agents
IBM watsonx Orchestrate
Notion Agents
Docebo Agents
Fin
Botpress
CodeWhale
Gemini Enterprise
HubSpot Agent Hub
innoGPT
Langdock
Lime Connect
SAP Joule Agents
ServiceNow AI Agents
UiPath Agents
Kiro Crew
Timon
AppuBot
MausBot
How we classify
The classification is editorial and follows fixed rules based on profile details. Three levels per axis, because the data does not support a finer scale.
- An account or app is enough: Start tasks after signing in or installing the app.
- Configure access or workflows: For example, connect apps or configure a workflow.
- Run it yourself: Your own machine or server, often your own model keys.
- Starts on a schedule or event: The agent can start new runs on its own. Individual actions may require approval.
- Works in the background with supervision: Scheduled or running tasks may need an open session or regular review.
- Starts on your request: You start each run yourself.
Run agents on your own infrastructure
The bars show which types of tools you can often run on your own computer or server. The profile explains which data goes to model providers or connected services.
- Self-hostable
- Hosted only or not stated
Show as table
| Self-hostable | Hosted only or not stated | |
|---|---|---|
| Agent orchestration | 12 | 3 |
| Agent workspaces | 11 | 5 |
| Coding agents | 30 | 14 |
| Agent builders | 12 | 19 |
| General-purpose agents | 25 | 45 |
| Specialist agents | 0 | 19 |
Use a model that fits your needs
Already have access to a model or want to run it locally? See where you can connect your own models and where the selection is set for you.
- Open choice
- Fixed choice across providers
- Own models only
- Models not stated
Show as table
| Open choice | Fixed choice across providers | Own models only | Models not stated | |
|---|---|---|---|---|
| Agent orchestration | 10 | 5 | 0 | 0 |
| Coding agents | 28 | 14 | 2 | 0 |
| Agent workspaces | 7 | 9 | 0 | 0 |
| Agent builders | 12 | 15 | 1 | 3 |
| General-purpose agents | 25 | 19 | 8 | 18 |
| Specialist agents | 0 | 9 | 0 | 10 |
Most frequently named model families
Show as table
| Model family | agents |
|---|---|
| OpenAI (GPT) | 107 |
| Anthropic (Claude) | 104 |
| Google (Gemini) | 82 |
| xAI (Grok) | 50 |
| DeepSeek | 45 |
| Moonshot (Kimi) | 39 |
| Z.ai (GLM) | 34 |
| Mistral | 32 |
| Alibaba (Qwen) | 26 |
| MiniMax | 23 |
The rating covers the generally available main access; betas and side tools are in the profile. An agent can belong to several tool types.
Check access and recovery before rollout
For recurring work, check which account the agent uses, who can intervene and how work resumes after an error. The questions with the highest share of gaps in documentation are listed first.
- Documented
- Documented with conditions
- Open in reviewed sources
Show as table
| Documented | Documented with conditions | Open in reviewed sources | |
|---|---|---|---|
| Sign-ins | 27 | 51 | 74 |
| Resuming after failures | 22 | 66 | 64 |
| Admin approval | 44 | 61 | 47 |
| Detecting errors | 46 | 61 | 45 |
| Human intervention | 71 | 52 | 29 |
| Files and working state | 78 | 46 | 28 |
| Account used | 100 | 41 | 11 |
| Starting further runs | 96 | 47 | 9 |
| Data access & team sharing | 88 | 61 | 3 |
Check what an agent remembers
When work spans conversations, check what the agent can retrieve later and what you can edit or delete. The chart shows how well these questions are documented across the catalogue.
- Documented
- Partly documented
- Not publicly established
Show as table
| Documented | Partly documented | Not publicly established | |
|---|---|---|---|
| Agent and channel boundaries | 80 | 67 | 5 |
| Storage and scope | 73 | 76 | 3 |
| Inspect, edit and delete | 70 | 66 | 16 |
| Selection and timing | 66 | 72 | 14 |
| Retrieval | 62 | 83 | 7 |
| Session reset and compaction | 26 | 113 | 13 |
Check availability in your country
An agent can be available in Europe while a specific feature is limited to certain countries or plans. The chart counts individual findings about access and features.
- Available
- With requirements
- Restricted
- Unclear
Show as table
| Available | With requirements | Restricted | Unclear | |
|---|---|---|---|---|
| Europe entries | 21 | 92 | 10 | 107 |
Estimate the cost of trying an agent
The price bands show the cheapest published access. Check the profile for model costs and usage limits relevant to your task.
Show as table
| Entry plan | agents |
|---|---|
| Free way in | 86 |
| Up to USD 20 a month | 13 |
| USD 21 to 50 | 10 |
| USD 51 to 200 | 2 |
| Over USD 200 | 1 |
| Usage-based only | 7 |
| On request only | 9 |
| Not clearly stated | 24 |
Explore projects with public source code
See which listed repositories receive the most attention on GitHub. Find licensing and usage requirements in each profile.
Show as table
| Agent | Stars |
|---|---|
| OpenClaw | 391,505 |
| Hermes Agent | 252,097 |
| OpenCode | 212,252 |
| n8n AI Agents | 206,759 |
| AutoGPT | 187.909 |
| Dify | 157,962 |
| Langflow | 156.355 |
| Codex | 128,271 |
| Pi | 120.243 |
| Browser Use | 117,341 |
| Gemini CLI | 107,256 |
| Paperclip | 98,947 |
Oldest figure as of: 9 October 2026
What has changed
First capture Changes appear here once a second capture exists.