OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Photo Agents

Photo Agents is a Python 3.10+ framework for photo-aware LLM agents that perceive a computer screen, reason through a tool-calling loop and act through local computer-use tools. It combines layered memory, including working, global, SOP and session-archive stores, with self-written skills that can be reused for recurring tasks. The package includes an LLM provider router, file and code-execution tools, browser automa

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Photo Agents — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Photo Agents when a task requires an LLM agent to interpret what is visible on a computer and perform actions through files, code, browsers or connected clients, especially when retaining local memory and reusable skills matters.

Who it is for

  • Developers building vision-grounded computer-use agents.
  • Builders creating desktop automation assistants that can retain task context.
  • Teams that want reusable skills and SOPs generated from agent work.
  • Self-hosted users who want local storage for screen-related state, memory and skills.

Documented uses

  • Perceive and act on a computer through an LLM-driven agent loop.
  • Run file and sandboxed code tasks using Python, PowerShell or bash tools.
  • Automate browser workflows through a Chrome DevTools Protocol bridge.
  • Store observations and session history in layered local memory.
  • Search and reuse skills or SOPs for recurring computer tasks.
Before you choose

Requirements & limits

The package is distributed as MIT-licensed open-source software. Starting the runtime requires a remotely validated Photo Agents API key and credentials for a supported LLM provider. Python 3.10 or newer is required; the README states testing on Windows, macOS and Linux.

  • The framework requires both a Photo Agents API key and an LLM provider credential before the agent can run.
  • The agent loop depends on a network-reachable LLM provider, while memory and skills are stored locally.
  • Browser automation depends on the documented Chrome DevTools Protocol bridge and its related configuration.
  • The package offers several frontends, but each uses the same API-key gate before launching.
  • The beta status means APIs may change before the 1.0 release.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install the Photo Agents runtime

Set up the Python package, Photo Agents API key and LLM credentials required to start the runtime.

Before you start

  • Python 3.10 or newer
  • A Photo Agents API key
  • A credential for a supported LLM provider
  • The Photo Agents repository or package source

Steps

  1. Install the package with pip, using the optional all-extra when the additional clients and integrations are needed.

  2. Provide the Photo Agents API key through PHOTOAGENTS_API_KEY, ~/.photoagents/config.json, or the interactive first-run prompt.

  3. Copy the credentials template to credentials.py or use the equivalent credentials.json form, then configure a provider entry.

  4. Start the package with the documented Python module command to open its terminal runtime.

Expected outcome

The local runtime is configured to start an interactive Photo Agents session.

Guide limitations

  • The agent loop requires a network-reachable LLM provider.
  • The runtime is gated by a remote-validated Photo Agents API key.
  • The project is beta and its APIs may change before 1.0.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
4 October 2026
Original website title
jmerelnyc/Photo-agents · Official homepage title · retained in its source language
How we review sourcesSuggest a correction