// MCP TOKEN COST

The Starfield MCP server costs 32,796 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 Starfield MCP server's 233 tool definitions consume 32,796 tokens — 16% of a 200k context window, and 17× the median MCP server (1,905 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS tiktoken o200k_base · rank #79 of 8,377 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 16%

1M WINDOW 3.3%

Corpus context: Starfield ranks #79 of 8,377 measured MCP servers by definition cost. The median is 1,905 tokens, p90 is 12,638, and the heaviest (Ainumbers Mcp Apps) is 327,608 — 164% 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 32,796 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 141 tokens per tool.

Tool	Category	Tokens	% of server
move_work_item	Write	882	2.7%
create_work_item	Write	818	2.5%
comment_on_work_item	Write	530	1.6%
list_my_work_items	Read	502	1.5%
flag_work_item	Destructive	480	1.5%
adrata_api_request	Execute	471	1.4%
set_work_item_tag	Write	471	1.4%
adrata_ai_tool_execute	Execute	386	1.2%
set_work_item_kind	Write	385	1.2%
attribute_partner_to_deal	Execute	354	1.1%
bulk_delete_buyer_groups	Destructive	350	1.1%
record_partner_consumption	Execute	335	1.0%
get_cloud_records	Read	309	0.9%
sloan_handoff	Write	309	0.9%
build_pursuit_command_center	Execute	305	0.9%
configure_sloan	Write	300	0.9%
update_buyer_group_member	Write	299	0.9%
log_interaction	Read	296	0.9%
schedule_meeting	Write	294	0.9%
recommend_deal_move	Write	293	0.9%
create_sequence	Write	285	0.9%
submit_adrata_command	Write	279	0.9%
remove_buyer_group_member	Destructive	270	0.8%
add_buyer_group_member	Write	268	0.8%
delete_buyer_group	Destructive	262	0.8%
list_customer_signals	Read	250	0.8%
update_buyer_group	Write	249	0.8%
request_provider_action_execution	Execute	247	0.8%
create_partner	Write	244	0.7%
create_person	Write	244	0.7%
approve_partner_attribution	Write	242	0.7%
create_buyer_group	Write	240	0.7%
sloan_status	Read	237	0.7%
paper_create_document	Write	236	0.7%
search_people	Read	234	0.7%
export_data	Write	229	0.7%
manage_custom_fields	Write	229	0.7%
request_deployment	Execute	222	0.7%
paper_create_share	Write	214	0.7%
move_pipeline_card	Write	210	0.6%
add_sequence_step	Write	209	0.6%
bulk_import	Write	209	0.6%
paper_update_document	Write	208	0.6%
schedule_demo	Write	202	0.6%
update_person	Write	202	0.6%
connect_workspace	Write	198	0.6%
create_action	Write	196	0.6%
get_installed_base	Read	195	0.6%
list_actions	Read	194	0.6%
get_work_item	Read	193	0.6%
paper_create_calendar_event	Write	192	0.6%
get_work_item_comments	Read	188	0.6%
paper_send_email	Write	188	0.6%
find_or_create_person	Read	179	0.5%
create_strategic_work_stream	Write	175	0.5%
get_account_read	Read	171	0.5%
get_company_invoices	Read	171	0.5%
create_email_account	Write	169	0.5%
save_memory	Write	169	0.5%
update_partner	Write	167	0.5%
request_workflow_deployment	Execute	166	0.5%
rank_paths_to_power	Write	166	0.5%
paper_archive_document	Write	165	0.5%
get_work_board_rollup	Read	164	0.5%
paper_delete_document	Destructive	162	0.5%
create_intro_request	Write	161	0.5%
morning_brief	Read	160	0.5%
paper_update_task	Write	157	0.5%
test_provider_credential	Read	153	0.5%
list_partners	Read	148	0.5%
create_executive_partner_meeting	Write	147	0.4%
dry_run_workflow	Execute	145	0.4%
paper_revoke_share	Destructive	144	0.4%
paper_create_task	Write	144	0.4%
recall	Read	142	0.4%
rank_partners_for_account	Read	140	0.4%
list_strategic_work_streams	Read	137	0.4%
inspect_provider_catalog	Read	134	0.4%
search_domains	Read	133	0.4%
validate_integration_manifest	Read	132	0.4%
create_webhook	Write	131	0.4%
list_partner_attributions	Read	129	0.4%
list_executive_partner_meetings	Read	128	0.4%
list_partner_consumption	Read	126	0.4%
track_conversion	Read	126	0.4%
update_webhook	Write	125	0.4%
draft_workflow	Execute	124	0.4%
enroll_contacts	Read	124	0.4%
score_company_icp	Read	124	0.4%
import_external_pipeline_members	Write	124	0.4%
list_email_accounts	Read	123	0.4%
import_external_companies	Write	123	0.4%
search_opportunities	Read	122	0.4%
create_agent_task	Write	122	0.4%
validate_extension_manifest	Read	121	0.4%
get_work_item_history	Read	120	0.4%
rank_people_by_icp	Write	120	0.4%
upgrade_account	Write	120	0.4%
update_action	Write	119	0.4%
create_opportunity	Write	118	0.4%
list_external_pipelines	Read	117	0.4%
list_provider_endpoints	Read	116	0.4%
search_emails	Read	116	0.4%
update_intro_request	Write	116	0.4%
update_opportunity	Write	116	0.4%
replay_workflow_run	Execute	115	0.4%
find_person	Read	115	0.4%
find_company	Read	114	0.3%
search_companies	Read	114	0.3%
validate_workflow_draft	Read	113	0.3%
list_external_companies	Read	112	0.3%
purchase_domain	Read	112	0.3%
get_work_board	Read	111	0.3%
list_external_pipeline_members	Read	109	0.3%
paper_list_documents	Read	108	0.3%
warmup_email	Execute	107	0.3%
find_or_create_company	Read	106	0.3%
get_demo_availability	Read	106	0.3%
update_company	Write	106	0.3%
list_intro_requests	Read	105	0.3%
check_inbox	Read	104	0.3%
describe_fields	Read	103	0.3%
get_icp_distribution	Read	103	0.3%
list_sequences_full	Read	103	0.3%
list_notes	Read	102	0.3%
enable_toolset	Write	102	0.3%
get_mcp_analytics	Read	101	0.3%
list_custom_fields	Read	101	0.3%
list_agent_tasks	Read	100	0.3%
list_meetings	Read	99	0.3%
create_company	Write	99	0.3%
list_company_lists	Read	96	0.3%
rank_companies_by_icp	Read	95	0.3%
search_leads	Read	94	0.3%
create_company_list	Write	94	0.3%
list_webhook_deliveries	Read	93	0.3%
get_pipeline_metrics	Read	92	0.3%
setup_domain	Write	92	0.3%
workspace_status	Read	91	0.3%
list_icp_profiles	Read	90	0.3%
verify_domain	Read	88	0.3%
list_campaigns	Read	87	0.3%
list_work_boards	Read	87	0.3%
create_note	Write	87	0.3%
switch_workspace	Write	87	0.3%
get_partner_revenue	Read	86	0.3%
get_company_opportunities	Read	85	0.3%
get_provider_connect_plan	Read	85	0.3%
get_sequence_analytics	Read	85	0.3%
list_webhooks	Read	85	0.3%
get_company_actions	Read	84	0.3%
get_company_people	Read	84	0.3%
get_email_health	Read	84	0.3%
get_webhook_delivery	Read	84	0.3%
list_buyer_groups	Read	83	0.3%
list_work_board_rollups	Read	82	0.3%
adrata_ai_tool_catalog	Read	81	0.2%
get_priority_pursuits	Read	81	0.2%
check_subscription	Read	80	0.2%
list_domains	Read	79	0.2%
activate_sequence	Write	79	0.2%
find_intro_path	Read	78	0.2%
replay_webhook_delivery	Read	78	0.2%
add_to_company_list	Write	78	0.2%
get_access_graph_projection	Read	76	0.2%
get_partner_consumption_rollup	Read	75	0.2%
pause_sequence	Read	75	0.2%
get_activity_summary	Read	74	0.2%
get_company_firmographics	Read	74	0.2%
forget	Destructive	73	0.2%
get_access_score	Read	73	0.2%
get_path_to_power	Read	73	0.2%
get_speedrun_list	Read	73	0.2%
list_toolsets	Read	72	0.2%
list_overdue_actions	Read	71	0.2%
paper_list_calendar_events	Read	70	0.2%
adrata_api_catalog	Read	69	0.2%
check_batch_import_status	Read	69	0.2%
list_today_actions	Read	69	0.2%
list_workspaces	Read	69	0.2%
who_am_i	Read	69	0.2%
list_users	Read	68	0.2%
get_partner	Read	66	0.2%
get_workspace_settings	Read	66	0.2%
list_sequences	Read	66	0.2%
adrata_desktop_app_audit	Read	65	0.2%
update_note	Write	65	0.2%
test_webhook	Read	63	0.2%
get_deal_authority	Read	61	0.2%
paper_app_audit	Read	61	0.2%
paper_list_tasks	Read	61	0.2%
disconnect_workspace	Write	61	0.2%
enrich_company	Read	60	0.2%
get_agent_task	Read	60	0.2%
get_buyer_group_members	Read	60	0.2%
complete_action	Write	60	0.2%
enrich_person	Read	59	0.2%
get_company_list	Read	59	0.2%
paper_get_company	Read	59	0.2%
get_company	Read	58	0.2%
get_meeting	Read	58	0.2%
get_person	Read	58	0.2%
delete_webhook	Destructive	57	0.2%
get_competitor_intel	Read	57	0.2%
get_intent_signals	Read	57	0.2%
get_opportunity	Read	57	0.2%
paper_get_document	Read	56	0.2%
paper_get_email_thread	Read	56	0.2%
get_buyer_group	Read	55	0.2%
get_intro_pipeline	Read	55	0.2%
paper_list_shares	Read	55	0.2%
cancel_agent_task	Destructive	54	0.2%
get_meeting_action_items	Read	54	0.2%
get_meeting_summary	Read	53	0.2%
paper_resolve_share	Write	53	0.2%
get_action	Read	51	0.2%
get_campaign	Read	51	0.2%
get_email	Read	51	0.2%
delete_partner	Destructive	50	0.2%
get_sequence	Read	50	0.2%
list_webhook_events	Read	50	0.2%
delete_action	Destructive	49	0.1%
delete_company	Destructive	49	0.1%
delete_opportunity	Destructive	49	0.1%
delete_person	Destructive	49	0.1%
get_forecast_data	Read	49	0.1%
get_network_stats	Read	49	0.1%
delete_note	Destructive	48	0.1%
get_user	Read	48	0.1%
count_emails	Read	47	0.1%
count_records	Read	45	0.1%
paper_list_email_threads	Read	43	0.1%
get_current_user	Read	42	0.1%

// CUT THE COST

Your agent uses a handful of these tools. It pays for all 233.

You don't need all 233 of those definitions in the window. PolicyLayer is an MCP gateway that sits in front of Starfield: 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 233 tools (no gateway)	32,796 tokens	—
3 granted tools	~422 tokens	−99%
5 granted tools	~704 tokens	−98%
10 granted tools	~1,408 tokens	−96%

The risk dividend: 15 of these 233 tools are critical-risk (destructive or financial) and cost 2,146 tokens (7% 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 Starfield — 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 STARFIELD TOKEN COST →

Instant setup, no code required.

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

//FAQ

Starfield token-cost questions.

How many tokens does the Starfield MCP server use?+

Its 233 tool definitions total 32,796 tokens — 16% 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 Starfield 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 Starfield's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Starfield 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 422 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 06-09-2026 from the PolicyLayer scan database over all 233 catalogued Starfield tools. Counts refresh with every site build.