The Document Generation MCP Benchmark is a 20-task comparison of two MCP servers — DocuQueue and Anvil — run through Claude 3.5 Sonnet agents on real document workflows, with Conga, Templafy, and Nutrient compared from public documentation.
We built DocuQueue, one of the two MCP servers measured here — so weigh these results with that in mind. We ran the identical task set against both providers and dated every run, and we have a stake in this outcome.
The margins are not uniform:
Read the tables, not just the headline.
We wanted to understand where agents get stuck when using MCPs to complete real document workflows. These results are a first step, and we plan to build on this benchmark by adding more harnesses and tasks covering a broader range of use cases. We use the results to improve DocuQueue's MCP.
We compare the performance of the MCPs across 20 tasks spanning common document workflows like template generation, form filling, and batch processing.
The tasks use anonymized queries derived from real user examples.
Performance results cover 20 tasks: 5 template generation, 4 form filling, 2 URL → PDF, 2 HTML render, 2 batch processing, and 5 MCP tool tasks.
For each (task, provider) combination we start a fresh Claude 3.5 Sonnet agent with a task brief, any supplied inputs, and the provider's MCP. The agent discovers tools and works through the task. We record its time and token usage, then verify the agent produced the expected outputs to pass the task (e.g. generating a PDF).
We use the publicly hosted MCP server for each provider. Tests were last run on: September 10, 2026.
Which tasks each measured MCP provider completed. Coverage means the agent produced valid output, not that it succeeded perfectly. A dash (—) marks a category the provider does not offer — see Feature Comparison. Conga, Templafy, and Nutrient were not run through the suite; they are compared from public documentation.
| MCP Provider | Template Gen 5 tasks |
Form Filling 4 tasks |
URL → PDF 2 tasks |
HTML Render 2 tasks |
Batch 2 tasks |
MCP Tools 5 tasks |
Total |
|---|---|---|---|---|---|---|---|
| DocuQueue | 5/5 | 4/4 | 2/2 | 2/2 | 2/2 | 5/5 | 20/20 |
| Anvil | 5/5 | 4/4 | 2/2 | 2/2 | 2/2 | 4/5 | 19/20 |
DocuQueue completed all 20 tasks across all 6 categories — the only full-coverage run in the benchmark. Anvil reached 19/20.
DocuQueue generates from Word templates, fills PDF forms, and batches documents from one MCP server. Anvil batches with limits; Templafy can't fill PDF forms; Nutrient doesn't generate from business templates; Conga's MCP gateway is undocumented.
Agents discovered DocuQueue's MCP tools at an average of 1,842 tokens — the lowest tool-discovery cost across our benchmark runs.
25 free credits to test DocuQueue's MCP with your agents. No credit card required.
Key capabilities across DocuQueue, Anvil, Conga, Templafy, and Nutrient.
| Feature | DocuQueue | Anvil | Conga | Templafy | Nutrient |
|---|---|---|---|---|---|
| PDF Generation | ✓ | ✓ | ✓ | ✓ | ✓ |
| DOCX Templates | ✓ | — | ✓ | ✓ | ✓ |
| PDF Form Filling | ✓ | ✓ | — | — | ✓ |
| Batch Processing | ✓ | Limited | ✓ | — | ✓ |
| MCP Server | ✓ | ✓ | Undocumented | ✓ | ✓ |
| Free Tier | 25 credits | Sandbox | — | 1 seat | 50 credits/mo |
| OAuth Support | ✓ | — | ✓ | ✓ | ✓ |
| Zapier/n8n Integration | ✓ | ✓ | ✓ | — | ✓ |
| Self-serve Sign-up | ✓ | ✓ | — | — | ✓ |
DocuQueue and Anvil reflect our measured runs. Conga, Templafy, and Nutrient were assessed from public documentation (October 2026) and were not run through the 20-task suite — Conga's MCP gateway exists but is undocumented. Features and pricing may change.
Dollar figures measure the agent's model-token usage, excluding MCP provider costs and subscriptions.
What your LLM provider charges to run the agent — not a bill from any MCP provider.
| Volume | DocuQueue | Anvil |
|---|---|---|
| Per document | $0.1692 | $0.2008 |
| 100 documents | $17 | $20 |
| 1,000 documents | $169 | $201 |
| 10,000 documents | $1,692 | $2,008 |
Blended across the 12 template, URL, HTML, and MCP-generation tasks in the suite; form-filling and batch runs are reported separately in the tabs below.
At 10,000 documents, token usage is $316 below Anvil ($1,692 vs $2,008). DocuQueue's API price is $9 for 250 documents — see MCP/API Pricing below.
| Task | DocuQueue | Anvil |
|---|---|---|
| Invoice from template | $0.12 | $0.14 |
| Contract with variables | $0.18 | $0.21 |
| Certificate generation | $0.14 | $0.17 |
| Business letter | $0.11 | $0.13 |
| Gift voucher | $0.13 | $0.16 |
| Average | $0.14 | $0.16 |
| Task | DocuQueue | Anvil |
|---|---|---|
| W-4 tax form | $0.24 | $0.28 |
| I-9 employment | $0.22 | $0.26 |
| 1040 tax return | $0.31 | $0.36 |
| Custom PDF form | $0.19 | $0.22 |
| Average | $0.24 | $0.28 |
| Task | DocuQueue | Anvil |
|---|---|---|
| Simple page to PDF | $0.15 | $0.18 |
| Complex layout to PDF | $0.21 | $0.25 |
| Average | $0.18 | $0.22 |
| Task | DocuQueue | Anvil |
|---|---|---|
| Basic HTML render | $0.13 | $0.16 |
| Styled HTML with CSS | $0.18 | $0.21 |
| Average | $0.16 | $0.19 |
| Task | DocuQueue | Anvil |
|---|---|---|
| Batch invoices (10) | $0.28 | $0.33 |
| Batch certificates (50) | $0.45 | $0.52 |
| Average | $0.37 | $0.43 |
| Task | DocuQueue | Anvil |
|---|---|---|
| Generate via MCP tool | $0.16 | $0.19 |
| Fill form via MCP | $0.21 | $0.25 |
| Batch via MCP | $0.32 | $0.38 |
| Multi-step workflow | $0.28 | $0.33 |
| Custom tool chain | $0.24 | $0.28 |
| Average | $0.24 | $0.29 |
What DocuQueue and Anvil actually charge per document for API and MCP generation — MCP tool calls draw from the same quotas as direct API calls. Conga, Templafy, and Nutrient price by contract or seat and are listed after the table. Model-token costs are excluded: that is your LLM provider's bill, not a charge from any provider here.
| Monthly volume | DocuQueue | Anvil |
|---|---|---|
| 250 documents / month | ||
| Total | $9 | $25 |
| Per document | $0.036 | $0.10 |
| 1,000 documents / month | ||
| Total | $29 | $100 |
| Per document | $0.029 | $0.10 |
| 5,000 documents / month | ||
| Total | $99 | $500 |
| Per document | $0.0198 | $0.10 |
| 10,000 documents / month | ||
| Total | $198* | $1,000 |
| Per document | $0.0198* | $0.10 |
At 250 documents DocuQueue is $9 versus Anvil's $25. At 10,000 documents it is $198 versus $1,000 — 64–80% below Anvil across volumes ($16–$802 less).
API prices checked October 2, 2026 from each provider's pricing page. DocuQueue published tiers ($9 / 250, $29 / 1,000, $99 / 5,000 documents); Anvil $0.10 per document (excludes its one-time 2,500 free credits). *DocuQueue at 10,000 documents assumes 2× the $99 / 5,000 tier — no higher tier is published. Free tiers excluded: DocuQueue 25 credits, Anvil 2,500 one-time. Excludes e-signatures, workflows, seats, and feature packs; prices may change.
These three price by contract or seat rather than self-serve per document, so they don't fit the volume table above.
| Provider | Pricing model | Entry price | Free tier |
|---|---|---|---|
| Conga | Enterprise, annual contract | From $1.35 / document (5,000 for $6,750 / year, AWS Marketplace) | None published |
| Templafy | Per seat | $40 / user / month (Team, up to 25 seats); Enterprise custom | 1 free seat + 14-day trial |
| Nutrient | Credit-based | 0.5 credits / document (PDF gen, form fill); 2 credits (DOCX template) | 50 credits / month |
Landscape pricing from each provider's public pricing pages and marketplace listings, checked October 3, 2026. Conga and Templafy are sales-led — neither publishes self-serve per-document pricing.
Sample tasks from each category with difficulty ratings and expected outcomes. Metrics shown are DocuQueue's own runs.
Each trial starts a fresh Claude 3.5 Sonnet agent with a task brief and the provider's MCP server. The agent discovers available tools and works through the task autonomously.
Model-token usage by the outer agent, including cached input. It excludes MCP provider subscription costs and document generation fees, so a cheaper agent run does not necessarily mean a cheaper total workflow.
Structural output: the PDF exists, contains the expected data, has correct formatting, and passes validation checks. It does not independently establish visual quality or business accuracy beyond the task requirements.
Monthly, and after any major API changes. Results are dated to show when tests were last executed, and we add new tasks as DocuQueue's capabilities expand.
Yes. All benchmark tasks are available in our GitHub repository. You can reproduce results using any MCP provider, and we encourage independent verification.
MCP (Model Context Protocol) is how AI agents actually discover and use tools, so testing via MCP shows real-world agent performance rather than just API capability. This reflects how 80%+ of DocuQueue usage happens.
The measured suite currently covers DocuQueue and Anvil — adding a provider takes harness work, and Conga's MCP gateway is undocumented. Conga, Templafy, and Nutrient are compared from public documentation instead, and every cell from that source is labeled in the table footnotes.
Blended across the 12 template, URL, HTML, and MCP-generation tasks, model-token cost averaged $0.1692 per document for DocuQueue and $0.2008 for Anvil. The four form-filling tasks averaged $0.24 and $0.28 respectively. These figures cover agent model tokens only — they exclude MCP provider subscriptions and per-document fees.
Yes — Templafy and Nutrient are marked as having an MCP server in the feature table above. Conga's MCP gateway exists but is undocumented, so its tool list could not be verified. None of the three were run through the 20-task suite; their rows come from public documentation checked in October 2026.
DocuQueue, Anvil, and Nutrient support PDF form filling; Conga and Templafy do not. The measured suite includes four form-filling tasks — W-4, I-9, 1040, and a custom PDF form — all completed by DocuQueue at an average model-token cost of $0.24 per form.
Each task runs 5 times per provider to account for variability, starting a fresh Claude 3.5 Sonnet agent each time. Results exclude retries and error-recovery attempts, and tests were last run on September 10, 2026.
Yes — DocuQueue includes 25 free credits, Anvil offers a one-time sandbox of 2,500 credits, Templafy includes 1 free seat, and Nutrient provides 50 credits per month; Conga has no free tier. The API pricing table on this page excludes these free tiers.