AnyCap Documentation

The capability runtime for AI agents. Create media, understand files, search the live web, and publish results through one CLI and one authentication flow.

Start in 30 seconds Install the CLI Connect through MCP

Add media and web tools to your AI agent

AnyCap gives Claude Code, Cursor, Codex, and other agents a consistent execution layer. Use the Agent Skill for natural-language workflows, the built-in MCP server for typed tool discovery, or call the CLI directly.

Start with the CLI installation guide if you need the executable, or connect an MCP host if your agent discovers tools through MCP. The commands below use your AnyCap account to run capabilities through the service.

Create

Generate and edit images, produce video, compose music, and build complete audio scenes.

Understand

Read screenshots, compare images, summarize video, and transcribe or analyze audio.

Research

Search the live web with structured results or grounded answers, then crawl pages to Markdown.

Deliver

Share files through Drive, publish static sites with Page, and hand projects between agents with Snapshot.

Choose the product surface before you implement

This site documents runnable commands and workflows. If you are still deciding which capability or agent integration fits the problem, use the product site for the evaluation first.

Explore capabilitiesCompare image, video, media understanding, web, delivery, and agent workflows before choosing an implementation path.

Find your agent integrationSee how AnyCap fits Claude Code, Codex, Cursor, Gemini CLI, and Manus.

Choose your integration

Use the CLI when you want explicit commands and JSON results in a script. Choose the Agent Skill when your coding agent should plan and run those commands for you. Choose MCP when your host needs named tools with structured inputs. These paths share the CLI login; you do not need a separate account for each integration.

CLIInstall one binary, authenticate once, discover the live model catalog, and call AnyCap directly from your terminal or agent workflow.

Agent SkillTeach your agent how to install, authenticate, choose models, call the CLI, and deliver results from plain-English requests.

Local MCP serverExpose 23 typed AnyCap tools to any host that supports local stdio MCP servers.

The shortest successful path

curl -fsSL https://anycap.ai/install.sh | sh
anycap login
anycap status

anycap --version verifies the installation. After login, use anycap status to check the account and service connection. For login problems, follow the authentication guide.

Then ask your agent to complete a task or run a capability directly. First list the current image models:

anycap image models

Replace MODEL_ID below with an ID from that list. Inspect its text-to-image schema before generating:

anycap image models MODEL_ID schema --mode text-to-image
anycap image generate \
  --prompt "minimal product hero, warm editorial light" \
  --model MODEL_ID \
  -o product-hero.png

On success, the CLI saves the image to product-hero.png and prints a JSON result. Continue with the image workflow to refine and share it, or use the CLI reference to discover commands and parameters.

Learn by outcome

Generate and share an imageCreate an image, inspect the JSON result, upload it, and return a browser-ready link.

Animate a local imageDiscover an image-to-video model and pass a local reference file safely.

Review with a humanCollect spatial annotations and narrated feedback, then turn it into an actionable revision.

Research the live webChoose between structured search, grounded search, and direct page crawling.