Compare
A SuperDoc alternative for AI agents
SuperDoc is an open-source DOCX editor with SDKs and an MCP server. Vespper is a hosted MCP built only for agents editing Word files. On our benchmark of 279 editing tasks, agents using Vespper passed about twice as many tasks as with SuperDoc MCP.
Benchmark
The headline numbers
| Metric | Vespper (Sol) | SuperDoc (Sol) | Vespper (Terra) | SuperDoc (Terra) |
|---|---|---|---|---|
| Pass rate | 81.4% | 41.7% | 78.9% | 41.7% |
| Median cost per task | $0.054 | $0.130 | $0.022 | $0.054 |
| Median latency | 14.6s | 16.8s | 13.7s | 11.5s |
| Median tool calls | 3 | 4 | 3 | 3 |
Models: GPT 5.6 Sol and GPT 5.6 Terra at medium reasoning. SuperDoc was slightly faster on GPT 5.6 Terra, but passed 41.7% of tasks against Vespper's 78.9%. We tested SuperDoc MCP v0.18.1. Its open, save, and close tools were hidden and handled for the agent, so setup cost it nothing. The full methodology covers the dataset, scoring, and all six solutions tested.
By task type
Where the gap comes from
The gap is widest on tables, headings, and tracked-change management. Those are the edits that depend on inherited styles and document structure.
| Task type (GPT 5.6 Sol pass rate) | Vespper DOCX MCP | SuperDoc MCP |
|---|---|---|
| Tables (n=90) | 77.8% | 25.8% |
| Lists (n=96) | 71.9% | 34.4% |
| Paragraphs (n=159) | 75.5% | 39.6% |
| Headings (n=28) | 60.7% | 17.9% |
| Tracked-change management (n=26) | 57.7% | 0.0% |
| Header/Footer (n=4) | 50.0% | 0.0% |
How they work
Operations vs a reconciler
SuperDoc
SuperDoc reads, renders and edits DOCX files through their OOXML parts. The same operations are available in its browser editor, headless SDKs, CLI and MCP server. Agents get a set of document operations, which means learning SuperDoc's tools and options before editing.
Vespper DOCX MCP
The agent reads the document as clean HTML and sends back the HTML it wants. A model trained for this task writes the matching OOXML for each changed block. Vespper then computes tracked changes and patches the original file. Two tools, and no Word concepts for the agent to learn.
| Vespper DOCX MCP | SuperDoc | |
|---|---|---|
| What it is | Hosted MCP for agents | DOCX editor, SDKs and MCP |
| How the agent edits | Edits clean HTML | Calls OOXML operations |
| Browser editor for your users | No | Yes |
| Open source, self-hosted | Enterprise self-hosting | Yes |
| Edits returned as Word tracked changes | Yes | Yes |
Choosing
When to use which
Pick SuperDoc if
- You need a Word editor embedded in your own web app.
- You want an open-source component you host yourself.
Pick Vespper if
- An agent does the editing and accuracy is what matters.
- Edits have to come back as clean Word tracked changes.
- Documents are long, with tables, lists, and inherited styles.
- You'd rather connect to a hosted endpoint than run infrastructure.
Switching
Point your agent at one MCP endpoint
Swap the SuperDoc MCP server for Vespper's hosted endpoint. The quickstart covers authentication and file handling.
from fastmcp import Client
client = Client(
"https://mcp.vespper.com/mcp",
auth="TOKEN",
)
async with client:
tools = await client.list_tools()
# Pass the tools to your agent in place of SuperDoc'sRun your own documents through it.
The free plan includes 500 edits a month. See pricing.