Deterministic tools, one connection, priced by the operation.
The pipeline tool runs compress → convert → resize in a single request. Fewer round-trips, far fewer tokens, no intermediate files for the agent to juggle.
Files are processed in memory and discarded instantly. Nothing stored, logged, or used for training — safe for private photos and documents.
Compress, convert, resize, crop, rotate, read metadata and compress PDFs — a single MCP server instead of stitching together many services.
1 credit per op (2 for AI vision, 1/step for pipelines), 50 free to start plus 25 free ops a day. Usage is prepaid, failed ops are refunded.
20+ image & PDF tools plus AI vision, behind one MCP server. Send a file, get a file.
Pipeline
Chain any steps in one call
Compress
Shrink images, keep the format
Convert
webp · avif · jpeg · png
Resize
Exact width/height, safe caps
Crop
Pixels or aspect ratio (16:9…)
Rotate · Flip
Any angle, mirror H/V
Adjust · Grayscale
Brightness, saturation, hue
Blur · Tint · Negate
Filters and color effects
Round · Border
Rounded corners, framed
Watermark
Overlay text, set opacity
PDF compress · merge
Slim, join, split, rotate
Image → PDF
Bundle images into one PDF
Metadata · PDF info
Read size, format, pages
20+ tools
One key, the whole toolbox
Preparing product photos usually means chaining compress, convert and resize per image — dozens of tool calls and tokens per file. With the pipeline tool the agent sends one request describing the whole chain and gets only the finished file back. Fewer round-trips, fewer tokens, no intermediate files. See ready-made pipeline flows →
Source
photo.jpg
Compress
step 1
Convert
to webp
Resize
1200px
Output
optimized.webp
One request · one billed op per step · only the final file comes back
An honest comparison. No single service is cheapest at everything, so here is exactly where each one wins and where omnikitbolt does.
| omnikitbolt | Cloudinary | TinyPNG | Gemini direct | |
|---|---|---|---|---|
| Scope | Image + PDF + AI vision | Image/video CDN | Compression | AI vision only |
| Connect | OAuth, no key to paste | API key | API key | API key + your code |
| Pricing | Pay-as-you-go | $89+/mo plan | Pay-as-you-go | Pay-as-you-go |
| Per op | $0.003–0.005 | ~$0.0004* | $0.009 (→$0.002) | ~$0.0008 (DIY) |
| Pipeline (chain in 1 call) | Yes | No | No | No |
| PDF tools | Yes | Limited | No | No |
| AI vision (describe/OCR/tags) | Yes | Add-on | No | Yes |
| Zero-retention | Yes | Stores assets | Uploads files | Depends on your setup |
| MCP-native | Yes | No | No | No |
* Cloudinary's per-transform price is lower but requires a monthly plan and is pooled with storage and bandwidth. TinyPNG is cheapest for pure compression at very high volume. omnikitbolt wins when an agent needs a bit of everything, on demand, with no setup — see the full comparison.
Add the MCP URL and the tools appear automatically. Prefer raw HTTP? Grab a key and call the REST API.
.svg){kind=small}
{
"mcpServers": {
"omnikitbolt": {
"url": "http://omnikitbolt.com/api/mcp"
}
}
}Add this to your MCP client (Claude, Cursor, …). On first use you'll sign in with Google and approve access — no API key to paste.
Files sent to the API and MCP are processed entirely in memory and discarded the moment the request finishes. Nothing is stored, nothing is logged, nothing is used to train models. Your agent can safely handle private photos, contracts and scans.
Received · processed in memory · discarded
Cheapest to run, safest by default, and it plugs into anything. No subscription, no per-seat, no lock-in.
1 credit per operation (1 per step in a pipeline). Buy credits once, spend them across the API, MCP and the website. The more you buy, the less each operation costs.
No expiry games · failed operations are refunded · cancel anytime (there's nothing to cancel — it's prepaid).
It is a Model Context Protocol (MCP) server that lets AI agents run omnikitbolt's 20+ image and PDF tools directly: compress, convert, resize, crop, rotate, flip, adjust, grayscale, blur, tint, negate, round corners, add borders, watermark, read metadata, compress and merge and split and rotate PDFs, turn images into a PDF, and chain any of these in a pipeline. The agent calls a tool, sends an image (as base64 or a URL) and gets the processed file back.
No key to paste. The MCP server uses OAuth 2.1: your agent opens a link, you sign in with Google and approve access, and the agent receives a token automatically. You get 25 free operations every day, plus 50 credits on your first authorization. If you prefer a raw integration, you can also generate an API key and call the REST API.
Instead of calling compress, then convert, then resize as three separate tool calls (three round-trips, three sets of tokens, and intermediate files to manage), the pipeline tool runs the whole chain in ONE call and returns only the final result. This is faster and dramatically cheaper in tokens for agents.
No. Every operation is zero-retention: files are processed in memory and discarded when the request ends. Nothing is written to disk, logged, or used for training. This makes it safe for an agent to process sensitive documents or private photos.
Free accounts get 25 free operations per day; Pro subscribers get a much larger daily allowance (500/day) so they can use the API/MCP without hitting the wall. Beyond that it is pay-per-use with prepaid credits: 1 credit per operation (a pipeline is 1 per step), or 2 credits for AI vision ops (describe, alt text, OCR, tags, filename). From about $0.003 per credit. Credits are shared across the omnikitbolt website and the API/MCP, and a failed operation is refunded automatically.
Any MCP-compatible client — Claude Desktop, Claude, Cursor, and other agent frameworks that support remote MCP servers over HTTP. Point the client at the omnikitbolt MCP URL and it discovers the tools automatically.
Yes. The full REST API is described in an OpenAPI 3.1 spec at https://www.omnikitbolt.com/openapi.json, and there is an LLM-friendly overview at https://www.omnikitbolt.com/llms.txt. Point your agent or codegen tool at either to integrate without guesswork.
Yes. Deterministic operations (compress, convert, resize, crop, rotate, PDF compress/split/rotate and more) have a keyless free trial: a few calls per day per IP with no API key and no account, so an agent can demonstrate the result in-chat before anyone signs up. After that, grab a free API key (25 ops/day) or connect over MCP. AI vision ops require a key.
Everything to evaluate and connect omnikitbolt as your agent's image toolbox.
Give your AI agent image tools
Set up Claude, Cursor or REST in minutes
Build image pipelines
Chain operations in one call, save tokens
Best image MCP servers 2026
Honest roundup of the ecosystem
Image API comparison for agents
omnikitbolt vs Cloudinary, TinyPNG, Gemini
omnikitbolt vs Cloudinary
Pay-as-you-go vs media CDN
omnikitbolt vs imgix
No bandwidth billing vs image CDN
Connect the omnikitbolt MCP server and get 25 free operations every day.
Get your key & connect