// MCP TOKEN COST

The Local MCP server costs 27,464 tokens before the first call.

Every request your agent makes carries every tool definition this server exposes — context your code, documents and conversation can't use, mostly for tools the agent never calls. You don't need them all in the window, and you don't have to pay for them.

QUICK ANSWER The Local MCP server's 234 tool definitions consume 27,464 tokens — 14% of a 200k context window, and 13× the median MCP server (2,142 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS tiktoken o200k_base · rank #59 of 5,402 measured servers · refreshed every build Method →

// CONTEXT WINDOW SHARE

What that costs 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 14%

1M WINDOW 2.7%

Corpus context: Local ranks #59 of 5,402 measured MCP servers by definition cost. The median is 2,142 tokens, p90 is 11,749, and the heaviest (Ainumbers Mcp Apps) is 161,750 — 81% of a 200k window on its own. New to this? See MCP token cost and context window in the glossary.

Token cost is one axis. See the risk picture across your whole stack →

// PER-TOOL BREAKDOWN

Where the 27,464 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 117 tokens per tool.

Tool	Category	Tokens	% of server
report_problem	Write	336	1.2%
chrome_read_tab	Read	286	1.0%
read_email	Read	274	1.0%
safari_read_tab	Read	272	1.0%
list_emails	Read	260	0.9%
create_draft	Write	258	0.9%
create_calendar_event	Write	257	0.9%
signal_friction	Write	254	0.9%
setup_install	Execute	253	0.9%
screen_record_start	Execute	242	0.9%
lmcp_welcome	Read	239	0.9%
run_terminal_command	Execute	239	0.9%
m365_create_event	Write	234	0.9%
update_calendar_event	Write	234	0.9%
list_calendar_events	Read	231	0.8%
servicenow_update_incident	Write	225	0.8%
stocks_get_chart	Read	222	0.8%
teams_send_message	Write	221	0.8%
update_reminder	Write	221	0.8%
ui_find_element	Read	219	0.8%
create_referral_invites	Write	217	0.8%
video_reframe	Read	209	0.8%
onedrive_read_file	Read	207	0.8%
video_export_gif	Write	199	0.7%
whatsapp_send_file	Write	195	0.7%
teams_call_history	Read	193	0.7%
gdrive_read_file	Read	193	0.7%
search_messages	Read	193	0.7%
list_email_folders	Read	191	0.7%
send_message	Write	185	0.7%
send_email	Write	184	0.7%
ui_wait_for_element	Execute	180	0.7%
safari_click	Execute	177	0.6%
m365_send_email	Write	173	0.6%
video_blur_region	Read	170	0.6%
m365_reply_email	Write	170	0.6%
signal_send_message	Write	168	0.6%
list_referral_candidates	Read	167	0.6%
web_login	Execute	167	0.6%
chrome_click	Execute	165	0.6%
video_trim	Write	164	0.6%
fs_search	Read	164	0.6%
zoom_read_transcript	Read	164	0.6%
reply_email	Write	161	0.6%
whatsapp_send_message	Write	161	0.6%
zalo_send_message	Write	161	0.6%
lmcp_state	Read	160	0.6%
move_email	Write	160	0.6%
request_feature	Write	160	0.6%
todo_create_task	Write	159	0.6%
onedrive_set_scope	Write	158	0.6%
nordvpn_servers	Read	157	0.6%
ui_click	Execute	157	0.6%
recipe_run	Execute	156	0.6%
get_weather	Read	155	0.6%
teams_send_channel_message	Write	155	0.6%
m365_list_emails	Read	153	0.6%
recipe_runs	Read	151	0.5%
web_show	Execute	151	0.5%
ui_menu_bar_click	Execute	151	0.5%
web_wait_for	Execute	150	0.5%
recipe_delete	Destructive	148	0.5%
zalo_read_messages	Read	148	0.5%
record_marker	Write	147	0.5%
notion_read_page	Read	145	0.5%
chrome_search_tabs	Read	144	0.5%
daily_brief	Read	144	0.5%
search_emails	Read	144	0.5%
screenshot_capture	Read	143	0.5%
fs_read	Read	142	0.5%
create_reminder	Write	141	0.5%
get_contact	Read	140	0.5%
list_windows	Read	140	0.5%
onedrive_list_files	Read	138	0.5%
get_audit_log	Read	137	0.5%
web_read	Read	137	0.5%
whatsapp_list_chats	Read	137	0.5%
excel_write_cell	Write	136	0.5%
whatsapp_search_messages	Read	135	0.5%
web_click	Execute	135	0.5%
m365_search_emails	Read	133	0.5%
update_local_mcp	Execute	131	0.5%
chrome_wait_for	Execute	130	0.5%
safari_wait_for	Execute	130	0.5%
rename_reminder_list	Write	130	0.5%
safari_search_tabs	Read	129	0.5%
run_diagnostics	Execute	127	0.5%
slack_list_channels	Read	127	0.5%
web_find	Read	127	0.5%
web_type	Write	126	0.5%
list_reminders	Read	125	0.5%
slack_search_messages	Read	125	0.5%
web_eval	Execute	125	0.5%
servicenow_create_incident	Write	125	0.5%
zalo_connect	Execute	122	0.4%
web_navigate	Execute	118	0.4%
gdrive_list_files	Read	118	0.4%
whatsapp_disconnect	Write	118	0.4%
save_attachment	Write	117	0.4%
servicenow_add_comment	Write	117	0.4%
m365_list_events	Read	116	0.4%
web_extract	Read	116	0.4%
fs_list	Read	115	0.4%
list_omnifocus_tasks	Read	115	0.4%
delete_reminder_list	Destructive	114	0.4%
signal_search_messages	Read	114	0.4%
get_m365_person	Read	113	0.4%
onedrive_search_files	Read	112	0.4%
connect_servicenow	Write	112	0.4%
slack_read_channel_messages	Read	111	0.4%
ui_keystroke	Execute	109	0.4%
video_concat	Write	109	0.4%
excel_create	Write	109	0.4%
todo_complete_task	Write	109	0.4%
whatsapp_read_messages	Read	108	0.4%
gdrive_search_files	Read	107	0.4%
gdrive_write_file	Write	107	0.4%
m365_read_email	Read	106	0.4%
notion_search	Read	106	0.4%
signal_read_messages	Read	106	0.4%
zoom_list_recordings	Read	106	0.4%
whatsapp_connect	Execute	106	0.4%
notion_read_database	Read	105	0.4%
search_m365_directory	Read	104	0.4%
todo_list_tasks	Read	104	0.4%
signal_connect	Read	104	0.4%
zalo_search_messages	Read	103	0.4%
update_self_diagnosis	Read	102	0.4%
search_contacts	Read	100	0.4%
create_note	Write	99	0.4%
word_create	Write	99	0.4%
notion_list_pages	Read	98	0.4%
zalo_disconnect	Write	98	0.4%
excel_read	Read	97	0.4%
create_omnifocus_task	Write	97	0.4%
stocks_get_quote	Read	96	0.3%
teams_list_channels	Read	96	0.3%
update_note	Write	96	0.3%
window_set_frame	Execute	95	0.3%
permissions_status	Read	94	0.3%
finder_search	Read	93	0.3%
notion_open_page	Execute	93	0.3%
servicenow_search_incidents	Read	92	0.3%
ppt_create	Write	92	0.3%
screen_record_stop	Execute	91	0.3%
signal_list_chats	Read	91	0.3%
finder_list	Read	90	0.3%
delete_reminder	Destructive	89	0.3%
read_messages	Read	89	0.3%
get_datetime	Read	88	0.3%
delete_calendar_event	Destructive	86	0.3%
media_probe	Read	86	0.3%
web_screenshot	Read	86	0.3%
web_session_close	Execute	86	0.3%
servicenow_list_my_incidents	Read	85	0.3%
complete_reminder	Write	85	0.3%
teams_read_channel_messages	Read	84	0.3%
ui_type	Execute	84	0.3%
word_append	Write	83	0.3%
zalo_list_chats	Read	83	0.3%
onedrive_write_file	Write	83	0.3%
list_m365_people_insights	Read	81	0.3%
servicenow_search_kb	Read	81	0.3%
m365_delete_event	Destructive	78	0.3%
create_email_folder	Write	77	0.3%
onedrive_move_file	Write	77	0.3%
gdrive_root	Read	76	0.3%
stocks_search_symbol	Read	75	0.3%
ui_get_element	Read	75	0.3%
onedrive_file_info	Read	74	0.3%
chrome_navigate	Execute	73	0.3%
safari_navigate	Execute	73	0.3%
m365_get_contact	Read	73	0.3%
chrome_type	Execute	73	0.3%
recipe_save	Write	73	0.3%
list_contacts	Read	72	0.3%
list_safari_bookmarks	Read	72	0.3%
teams_read_chat_messages	Read	72	0.3%
create_reminder_list	Write	72	0.3%
list_displays	Read	71	0.3%
list_omnifocus_folders	Read	69	0.3%
safari_type	Execute	68	0.2%
gdrive_file_info	Read	67	0.2%
pdf_read	Read	67	0.2%
onedrive_delete_file	Destructive	66	0.2%
servicenow_get_incident	Read	66	0.2%
slack_list_workspaces	Read	66	0.2%
window_focus	Execute	64	0.2%
outlook_diagnose	Read	64	0.2%
chrome_evaluate_js	Execute	63	0.2%
chrome_query_selector_all	Read	62	0.2%
list_email_accounts	Read	62	0.2%
safari_query_selector_all	Execute	62	0.2%
safari_evaluate_js	Execute	61	0.2%
complete_omnifocus_task	Write	61	0.2%
m365_search_contacts	Read	60	0.2%
chrome_setup_check	Read	60	0.2%
connect_m365_account	Write	58	0.2%
m365_list_contacts	Read	57	0.2%
chrome_fill_form	Execute	57	0.2%
safari_setup_check	Read	56	0.2%
search_omnifocus_tasks	Read	55	0.2%
list_calendar_names	Read	54	0.2%
list_reminder_lists	Read	54	0.2%
ppt_read	Read	54	0.2%
teams_list_chats	Read	53	0.2%
todo_list_lists	Read	53	0.2%
notion_list_databases	Read	52	0.2%
word_read	Read	52	0.2%
safari_fill_form	Execute	52	0.2%
nordvpn_diagnose	Execute	51	0.2%
screen_record_status	Read	51	0.2%
list_omnifocus_projects	Read	50	0.2%
zalo_diagnose	Read	50	0.2%
nordvpn_status	Read	49	0.2%
search_notes	Read	49	0.2%
list_message_chats	Read	48	0.2%
list_notes	Read	48	0.2%
recipe_list	Read	48	0.2%
notion_list_workspaces	Read	45	0.2%
read_note	Read	45	0.2%
recipe_get	Read	45	0.2%
web_session_list	Read	45	0.2%
chrome_go_back	Execute	43	0.2%
safari_go_back	Execute	43	0.2%
chrome_list_tabs	Read	41	0.1%
safari_list_tabs	Read	41	0.1%
teams_list_teams	Read	36	0.1%
disconnect_m365_account	Write	36	0.1%
get_config	Read	35	0.1%
onedrive_root	Read	34	0.1%
disconnect_servicenow	Write	34	0.1%
list_omnifocus_tags	Read	32	0.1%
list_accounts	Read	31	0.1%

// CUT THE COST

Your agent uses a handful of these tools. It pays for all 234.

You don't need all 234 of those definitions in the window. PolicyLayer is an MCP gateway that sits in front of Local: only the tools you grant are exposed to the agent, the rest never load. A smaller window means a sharper agent — less noise when it picks a tool — and every request costs less:

Grant scope	Definition cost	Reduction
All 234 tools (no gateway)	27,464 tokens	—
3 granted tools	~352 tokens	−99%
5 granted tools	~587 tokens	−98%
10 granted tools	~1,174 tokens	−96%

The risk dividend: 6 of these 234 tools are critical-risk (destructive or financial) and cost 581 tokens (2% of the definition load). Block them — the recommended starter policy — and you reclaim that context before tuning anything else.

Create a free account and register Local — nothing to install.
Grant only the tools you use — ungranted definitions never enter the context window.
Point your MCP client (Claude, Cursor, anything) at your gateway URL.

CUT LOCAL TOKEN COST →

Instant setup, no code required.

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

//FAQ

Local token-cost questions.

How many tokens does the Local MCP server use?+

Its 234 tool definitions total 27,464 tokens — 14% 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 Local 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 Local's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Local 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 352 tokens, a 99% 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 23-07-2026 from the PolicyLayer scan database over all 234 catalogued Local tools. Counts refresh with every site build.