Best AI Vision Tools for Agents (2026)
Computer vision tools, image and video agents, and vision MCP servers that let AI agents see and understand images and video. Below are the top 20 of 21 matching tools in the AgentScout directory, ranked.
How we ranked: Ranked by GitHub stars as recorded on each AgentScout listing (link checks 2026-10-18). Tools without a tracked public repo come after, in alphabetical order. Stars measure developer popularity, not quality or fit for your job, so read each listing before you choose. No paid placements.

1. CVAT AI Agents
CVAT feature that runs your own detection, segmentation or tracking model as an agent on your infrastructure and feeds its auto-annotations back into CVAT tasks.
Pricing (as stated): Freemium · Active

2. Autodistill
Roboflow's open-source library that uses large foundation models to auto-label images and then trains small, deployable detection or classification models with no manual labeling.
Pricing (as stated): Free, open source (Apache-2.0) · Active

3. NVIDIA VSS Blueprint
NVIDIA reference blueprint for video AI agents that search, summarize, answer questions and raise alerts over live or recorded video, with agent skills that assemble deployments from one prompt.
Pricing (as stated): Free, open source (Apache-2.0) · Active

4. VideoDB Director
Open-source framework of 20+ video agents that search, summarize, clip, edit and stream video from natural-language commands, built on VideoDB.
Pricing (as stated): Free, open source (MIT); runs on the VideoDB API · Active

5. ImageSorcery MCP
Local MCP server that gives AI assistants object detection, OCR, cropping, background removal and drawing tools, running computer vision models on your own machine.
Pricing (as stated): Free, open source (MIT) · Active

6. NVIDIA TAO Agent Skills
NVIDIA's vision-model post-training toolkit, now packaged as agent skills so Claude Code or Codex can fine-tune, optimize and export vision foundation models from plain-English prompts.
Pricing (as stated): Free, open source (Apache-2.0) · Active

7. Telekinesis Agentic Skill Library
Python library of 250+ perception, vision and robotics skills plus a Physical AI agent that turns natural-language prompts into reviewable robot code.
Pricing (as stated): Freemium · Active

8. Label Studio MCP Server
HumanSignal's MCP server that lets AI assistants create Label Studio labeling projects, import tasks, read annotations and add model predictions.
Pricing (as stated): Free, open source (Apache-2.0) · Active

9. Ultralytics Agents
Visual canvas in Ultralytics Platform that chains YOLO models, vision-language models, conditions and actions like Slack alerts, plus official agent skills that let coding agents train, tune and export YOLO models.
Pricing (as stated): Freemium · Early access

10. Bubbaloop
Kornia's open-source single-binary agent runtime with an MCP server that orchestrates cameras, sensors and robots on Jetson, Raspberry Pi or Linux through chat.
Pricing (as stated): Free, open source (Apache-2.0) · Early

11. FiftyOne MCP Server
Voxel51's MCP server and skills that let AI agents load, explore, curate and evaluate computer vision datasets and run FiftyOne operators and the App from natural language.
Pricing (as stated): Free, open source (Apache-2.0) · Active

12. Encord Agents
Encord's agents that plug foundation models or your own models into labeling workflows to pre-label, segment, track, route and QA image, video and multimodal data, with a no-code catalog and an open-source SDK.
Pricing (as stated): Paid · Active

13. Picsellia MCP Server
Official MCP server for Picsellia's computer vision MLOps platform, giving AI assistants access to datasets, experiments, models, deployments and monitoring.
Pricing (as stated): Paid · Active

14. Encord Merlin
visionPaidEncord's beta MCP server and in-app agent that creates annotation projects, ontologies and workflows and reports label-quality analytics from Claude, Codex or other AI tools.
Pricing (as stated): Paid · Beta

15. EyePop.ai
visionFreemiumSDK-first vision platform where you define what to detect and it builds detection, tracking, OCR and VLM pipelines, with a Video Agent and agent skills, deployable to cloud, edge or on-prem.
Pricing (as stated): Freemium · Active

16. Groundlight
visionSee siteTurns any camera into a vision sensor: you ask a yes/no question in plain English and Groundlight builds and tunes the model, backed by human review, with edge deployment.
Pricing (as stated): See site · Active

17. Roboflow Agent
visionPaidComputer-vision agent that picks models and builds/deploys vision apps; plugs into Claude Code, Codex, Cursor via MCP.
Pricing (as stated): Paid platform (see site) · Launched Oct 14

18. TwelveLabs Jockey MCP
visionFreemiumTwelveLabs' hosted MCP server for Jockey, a video-understanding agent that searches, tracks entities and answers questions across your videos and images with clip citations.
Pricing (as stated): Freemium · Active

19. Valossa Assistant MCP
visionFreemiumEU-based video AI MCP server that lets Claude and other agents upload and search videos and pull transcripts, visual scenes, soundbites and ready-made clips.
Pricing (as stated): Freemium · Active

20. Viso Now
visionFreemiumDescribe what a camera should watch for in plain language and it builds a live vision app with detection logic, alerts and integrations, no datasets or training needed.
Pricing (as stated): Freemium · Active
Browse by category
Related: Images and design · Free AI image generator
More best-of lists: All best lists · Best AI Coding Agents · Best AI Agent Frameworks · Best AI Browser Agents · Best Open-Source MCP Servers · Most-Starred Open-Source AI Agent Projects · Best Self-Hosted AI Agents · Best Free AI Agents (Free Plans) · Best No-Code AI Agent Builders · Best AI Voice Agent Platforms · Best AI Agent Observability & Eval Tools · Best Personal AI Agents · Best AI Agents for Business Teams · Best Crypto AI Agents · Best AI Productivity Agents
Facts come from our directory listings; pricing is as each maker states it and GitHub stars are as recorded on each listing. Check official sites before you choose.