Home / Token cost / Useful AI

The Useful AI MCP server costs 2,145 tokens before the first call.

Connect Useful AI and its 27 tool definitions are loaded into the model's context on every request — 1.1% of a 200k window spent before your agent does anything.

QUICK ANSWER The Useful AI MCP server's tool definitions consume 2,145 tokens — around the median MCP server (1,905 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS 27 tools · 2,145 tokens · 1.1% of 200k · 0.2% of 1M Method →

What that buys before your agent starts working.

Tool definitions are overhead: they occupy context on every request and compete with your code, documents and conversation history for the same window.

200K WINDOW 1.1%
1M WINDOW 0.2%

Corpus context: Useful AI ranks #1526 of 3,213 measured MCP servers by definition cost. The median is 1,905 tokens, p90 is 7,952, and the heaviest (Fusionauth) is 183,337 — 92% of a 200k window on its own.

Where the 2,145 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 79 tokens per tool.

ToolCategoryTokens% of server
useful_recurrence_rule_expander Read 117 5.5%
useful_income_narrative_batch_normalizer Read 108 5.0%
useful_suggest Read 106 4.9%
useful_dispatch Read 103 4.8%
useful Read 102 4.8%
useful_batch_slug_generator Read 101 4.7%
useful_color_contrast_fixer Read 97 4.5%
suggest Read 95 4.4%
useful_ndc_batch_normalizer_and_formulary_validator Read 90 4.2%
useful_cross_source_field_reconciler Read 89 4.1%
useful_phone_number_parser Read 86 4.0%
useful_batch_record_normalizer Read 85 4.0%
useful_prompt_drift_detector Read 85 4.0%
useful_address_string_parser Read 84 3.9%
useful_mock_data_generator Read 84 3.9%
useful_royalty_escalation_clause_parser Read 77 3.6%
useful_contact_entry_parser Read 72 3.4%
useful_phone_number_normalizer Read 71 3.3%
useful_xml_schema_validator Read 62 2.9%
useful_password_policy_validator Read 59 2.8%
useful_food_temperature_inspection_parser Read 58 2.7%
useful_certificate_chain_inspector Read 55 2.6%
useful_css_inline_tool Read 55 2.6%
useful_logit_lens_simulator Read 54 2.5%
useful_temporal_expression_normalizer Execute 51 2.4%
useful_prompt_x_ray Read 51 2.4%
useful_token_attention_heatmap Read 48 2.2%

Most agents use a handful of these tools. They pay for all 27.

A PolicyLayer grant exposes only the tools you allow — ungranted definitions are filtered out of the tool list, so they never enter the context window. Estimates below assume typical-weight tools (79 tokens each).

Grant scopeDefinition costReduction
All 27 tools (no gateway) 2,145 tokens
3 granted tools ~238 tokens −89%
5 granted tools ~397 tokens −81%
10 granted tools ~794 tokens −63%

Useful AI token-cost questions.

How many tokens does the Useful AI MCP server use?+

Its 27 tool definitions total 2,145 tokens — 1.1% of a 200k context window — measured with tiktoken o200k_base over the serialised tools/list payload. Exact counts vary slightly by client and model.

Why does Useful AI consume tokens before I send a message?+

MCP clients load every connected server's tool definitions — name, description, and input schema — into the model's context so it knows what it can call. That payload is charged against your context window on every request, whether or not a tool is used.

How do I reduce Useful AI's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Useful AI to only the tools you allow — ungranted definitions are filtered out of the tool list, so they never enter the context window. A grant of 3 typical tools costs roughly 238 tokens, a 89% reduction.

Does deferred tool loading fix this?+

Partially, in some clients. Claude Code defers MCP tool schemas behind a tool-search step by default, and VS Code has experimental grouping — but you still pay tokens per search and reload, and Cursor, Windsurf and Gemini CLI load definitions upfront. Reducing the exposed tool set cuts the cost in every client.

How these numbers were measured.

01
Serialisation

Each tool is serialised as a tools/list entry — name, description, input schema — from the schemas in the PolicyLayer scan database. Clients differ slightly in framing, so treat counts as close estimates.

02
Tokeniser

tiktoken o200k_base (GPT-4o/o-series). Anthropic's current tokeniser isn't published, so Claude's exact counts will differ; for English text and JSON schemas the totals are close enough to treat these as estimates.

03
Deferred loading

Some clients now defer schema loading (Claude Code's tool search; VS Code experimental grouping). You still pay per search and reload — and Cursor, Windsurf and Gemini CLI load everything upfront.

Computed 07-06-2026 from the PolicyLayer scan database over all 27 catalogued Useful AI tools. Counts refresh with every site build.

Expose only the tools you use — the rest never enter your context.

A PolicyLayer grant scopes Useful AI to the tools you actually allow. Ungranted definitions never load, and every call that does run is checked against policy first.

Free to start. No card required.

4,600+ MCP servers and 31,000+ tools scanned and risk-classified.

// GET IN TOUCH

Have a question or want to learn more? Send us a message.

Message sent.

We'll get back to you soon.