// MCP TOKEN COST

The Agda MCP server costs 9,523 tokens before the first call.

Connect Agda and its 72 tool definitions are loaded into the model's context on every request — 4.8% of a 200k window spent before your agent does anything.

QUICK ANSWER The Agda MCP server's tool definitions consume 9,523 tokens — 5.0× the median MCP server (1,905 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS 72 tools · 9,523 tokens · 4.8% of 200k · 1.0% of 1M Method →

// CONTEXT WINDOW SHARE

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 4.8%

1M WINDOW 1.0%

Corpus context: Agda ranks #184 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.

// PER-TOOL BREAKDOWN

Where the 9,523 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 132 tokens per tool.

Tool	Category	Tokens	% of server
agda_load	Read	386	4.1%
agda_typecheck	Read	380	4.0%
agda_bug_report_bundle	Read	371	3.9%
agda_bug_report_update_bundle	Write	368	3.9%
agda_apply_edit	Write	315	3.3%
agda_list_modules	Read	278	2.9%
agda_impact	Execute	266	2.8%
agda_auto	Read	228	2.4%
agda_term_search	Read	205	2.2%
agda_read_module	Read	190	2.0%
agda_cache_info	Read	188	2.0%
agda_compile	Execute	185	1.9%
agda_apply_rename	Write	178	1.9%
agda_refine	Read	171	1.8%
agda_case_split	Read	168	1.8%
agda_give	Read	166	1.7%
agda_add_missing_clauses	Write	165	1.7%
agda_backend_hole	Write	165	1.7%
agda_load_no_metas	Read	162	1.7%
agda_intro	Read	156	1.6%
agda_refine_exact	Read	156	1.6%
agda_suggest_import	Write	156	1.6%
agda_solve_one	Read	155	1.6%
agda_infer_fixity_conflicts	Read	148	1.6%
agda_find_clash_source	Read	143	1.5%
agda_postulate_closure	Read	143	1.5%
agda_effective_options	Read	132	1.4%
agda_check_postulates	Read	125	1.3%
agda_load_highlighting_info	Read	120	1.3%
agda_search_definitions	Read	120	1.3%
agda_token_highlighting	Read	119	1.2%
agda_infer	Read	116	1.2%
agda_show_module	Read	116	1.2%
agda_stdlib_migration_map	Read	114	1.2%
agda_goal_type_context_check	Write	114	1.2%
agda_helper_function	Read	113	1.2%
agda_why_in_scope	Read	113	1.2%
agda_goal_type_context_infer	Write	111	1.2%
agda_compute	Read	110	1.2%
agda_backend_top	Write	109	1.1%
agda_elaborate	Read	106	1.1%
agda_solve_all	Read	102	1.1%
agda_bulk_status	Read	101	1.1%
agda_highlight	Read	101	1.1%
agda_goal_analysis	Read	100	1.1%
agda_verify_builtin	Read	100	1.1%
agda_goal_type	Write	97	1.0%
agda_goal	Read	94	1.0%
agda_project_config	Read	94	1.0%
agda_context	Read	92	1.0%
agda_project_progress	Read	85	0.9%
agda_search_about	Read	81	0.9%
agda_triage_error	Read	81	0.9%
agda_proof_status	Read	80	0.8%
agda_metas	Read	72	0.8%
agda_show_irrelevant_args	Read	71	0.7%
agda_show_implicit_args	Read	70	0.7%
agda_tools_catalog	Read	70	0.7%
agda_protocol_parity	Read	65	0.7%
agda_builtin_migration_map	Read	62	0.7%
agda_reload	Read	61	0.6%
agda_goal_catalog	Read	60	0.6%
agda_session_snapshot	Read	59	0.6%
agda_tool_recommend	Read	52	0.5%
agda_auto_all	Read	50	0.5%
agda_toggle_irrelevant_args	Read	48	0.5%
agda_toggle_implicit_args	Read	47	0.5%
agda_constraints	Write	46	0.5%
agda_show_version	Read	42	0.4%
agda_session_status	Read	40	0.4%
agda_exit	Write	39	0.4%
agda_abort	Write	31	0.3%

// SCOPED GRANT

Most agents use a handful of these tools. They pay for all 72.

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 (132 tokens each).

Grant scope	Definition cost	Reduction
All 72 tools (no gateway)	9,523 tokens	—
3 granted tools	~397 tokens	−96%
5 granted tools	~661 tokens	−93%
10 granted tools	~1,323 tokens	−86%

Model your own stack in the token-cost calculator, or see the Agda policy for what a sensible grant looks like.

//FAQ

Agda token-cost questions.

How many tokens does the Agda MCP server use?+

Its 72 tool definitions total 9,523 tokens — 4.8% 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 Agda 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 Agda's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Agda 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 397 tokens, a 96% 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.

// METHOD & REVIEW

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 72 catalogued Agda tools. Counts refresh with every site build.