// MCP TOKEN COST

The Portkey Admin MCP server costs 46,570 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 Portkey Admin MCP server's 150 tool definitions consume 46,570 tokens — 23% of a 200k context window, and 25× the median MCP server (1,900 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS tiktoken o200k_base · rank #13 of 3,165 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 23%

1M WINDOW 4.7%

Corpus context: Portkey Admin ranks #13 of 3,165 measured MCP servers by definition cost. The median is 1,900 tokens, p90 is 7,952, and the heaviest (Fusionauth) is 183,337 — 92% 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 46,570 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 310 tokens per tool.

Tool	Category	Tokens	% of server
get_analytics_group_metadata	Read	1,170	2.5%
get_analytics_group_models	Read	1,142	2.5%
get_analytics_group_users	Read	1,140	2.4%
get_latency_analytics	Read	1,101	2.4%
get_users_analytics	Read	1,097	2.4%
get_cache_hit_rate	Read	1,091	2.3%
get_token_analytics	Read	1,091	2.3%
get_error_analytics	Read	1,090	2.3%
get_request_analytics	Read	1,090	2.3%
get_feedback_analytics	Read	1,089	2.3%
get_cost_analytics	Read	1,088	2.3%
get_error_stacks_analytics	Read	1,083	2.3%
get_cache_hit_latency	Read	1,082	2.3%
get_error_status_codes_analytics	Read	1,080	2.3%
get_error_rate_analytics	Read	1,079	2.3%
get_rescued_requests_analytics	Read	1,079	2.3%
get_feedback_models_analytics	Read	1,078	2.3%
get_user_requests_analytics	Read	1,078	2.3%
get_feedback_weighted_analytics	Read	1,076	2.3%
get_feedback_scores_analytics	Read	1,073	2.3%
create_prompt	Write	836	1.8%
migrate_prompt	Write	782	1.7%
update_prompt	Write	710	1.5%
run_prompt_completion	Execute	497	1.1%
create_guardrail	Write	482	1.0%
update_guardrail	Write	476	1.0%
create_api_key	Write	463	1.0%
invite_user	Write	461	1.0%
insert_log	Write	459	1.0%
update_api_key	Write	451	1.0%
create_log_export	Write	418	0.9%
create_provider	Write	416	0.9%
create_rate_limit	Write	400	0.9%
create_usage_limit	Write	400	0.9%
create_integration	Write	391	0.8%
update_provider	Write	390	0.8%
list_audit_logs	Read	347	0.7%
get_user_stats	Read	341	0.7%
render_prompt	Read	325	0.7%
create_virtual_key	Write	324	0.7%
create_config	Write	321	0.7%
update_integration	Write	320	0.7%
create_mcp_integration	Write	315	0.7%
update_log_export	Write	295	0.6%
update_config	Write	285	0.6%
add_workspace_member	Write	258	0.6%
update_integration_workspaces	Write	253	0.5%
create_feedback	Write	227	0.5%
update_mcp_integration	Write	227	0.5%
update_usage_limit	Write	226	0.5%
promote_prompt	Write	224	0.5%
update_virtual_key	Write	218	0.5%
update_workspace	Write	213	0.5%
update_integration_models	Write	198	0.4%
update_mcp_integration_capabilities	Write	192	0.4%
validate_completion_metadata	Read	190	0.4%
create_prompt_label	Write	189	0.4%
delete_api_key	Destructive	188	0.4%
list_integrations	Read	188	0.4%
get_api_key	Read	186	0.4%
update_mcp_server_capabilities	Write	182	0.4%
list_prompts	Read	180	0.4%
update_prompt_label	Write	179	0.4%
create_workspace	Write	177	0.4%
update_feedback	Write	176	0.4%
update_rate_limit	Write	176	0.4%
list_providers	Read	166	0.4%
list_guardrails	Read	165	0.4%
list_prompt_labels	Read	165	0.4%
update_mcp_integration_workspaces	Write	164	0.4%
update_mcp_server_user_access	Write	163	0.4%
list_integration_workspaces	Read	160	0.3%
create_mcp_server	Write	160	0.3%
update_prompt_partial	Write	159	0.3%
list_api_keys	Read	153	0.3%
create_prompt_partial	Write	152	0.3%
list_collections	Read	151	0.3%
list_integration_models	Read	149	0.3%
list_mcp_integrations	Read	147	0.3%
publish_prompt	Write	147	0.3%
list_mcp_server_user_access	Read	145	0.3%
list_all_users	Read	144	0.3%
list_mcp_server_capabilities	Read	144	0.3%
update_prompt_version	Write	141	0.3%
list_mcp_servers	Read	137	0.3%
publish_partial	Write	136	0.3%
update_user	Write	136	0.3%
list_virtual_keys	Read	133	0.3%
list_user_invites	Read	132	0.3%
update_workspace_member	Write	132	0.3%
list_configs	Read	129	0.3%
list_workspaces	Read	126	0.3%
create_collection	Write	121	0.3%
resend_user_invite	Write	121	0.3%
get_guardrail	Read	120	0.3%
get_prompt_label	Read	118	0.3%
cancel_log_export	Destructive	116	0.2%
delete_provider	Destructive	116	0.2%
update_mcp_server	Write	116	0.2%
update_collection	Write	112	0.2%
get_user	Read	111	0.2%
get_prompt	Read	110	0.2%
get_workspace	Read	110	0.2%
delete_integration_model	Destructive	108	0.2%
list_prompt_partials	Read	108	0.2%
get_config	Read	107	0.2%
reset_usage_limit_entity	Destructive	106	0.2%
get_prompt_version	Read	105	0.2%
get_provider	Read	105	0.2%
start_log_export	Execute	103	0.2%
get_integration	Read	103	0.2%
get_mcp_integration	Read	102	0.2%
get_workspace_member	Read	102	0.2%
list_mcp_integration_capabilities	Read	102	0.2%
remove_workspace_member	Destructive	100	0.2%
list_log_exports	Read	100	0.2%
get_mcp_integration_metadata	Read	99	0.2%
list_prompt_versions	Read	99	0.2%
get_log_export	Read	98	0.2%
delete_prompt	Destructive	97	0.2%
get_mcp_server	Read	97	0.2%
list_partial_versions	Read	96	0.2%
list_workspace_members	Read	96	0.2%
download_log_export	Read	95	0.2%
get_prompt_partial	Read	95	0.2%
list_mcp_integration_workspaces	Read	95	0.2%
list_rate_limits	Read	95	0.2%
delete_user	Destructive	93	0.2%
test_mcp_server	Read	93	0.2%
list_usage_limits	Read	92	0.2%
delete_prompt_partial	Destructive	91	0.2%
get_virtual_key	Read	91	0.2%
delete_mcp_integration	Destructive	90	0.2%
get_collection	Read	90	0.2%
delete_guardrail	Destructive	89	0.2%
delete_virtual_key	Destructive	89	0.2%
list_usage_limit_entities	Read	89	0.2%
delete_mcp_server	Destructive	88	0.2%
delete_usage_limit	Destructive	88	0.2%
get_user_invite	Read	88	0.2%
get_usage_limit	Read	87	0.2%
list_config_versions	Read	87	0.2%
delete_collection	Destructive	85	0.2%
delete_prompt_label	Destructive	85	0.2%
delete_integration	Destructive	84	0.2%
get_rate_limit	Read	82	0.2%
delete_user_invite	Destructive	81	0.2%
delete_config	Destructive	80	0.2%
delete_rate_limit	Destructive	80	0.2%
delete_workspace	Destructive	79	0.2%

// CUT THE COST

Your agent uses a handful of these tools. It pays for all 150.

You don't need all 150 of those definitions in the window. PolicyLayer is an MCP gateway that sits in front of Portkey Admin: 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 150 tools (no gateway)	46,570 tokens	—
3 granted tools	~931 tokens	−98%
5 granted tools	~1,552 tokens	−97%
10 granted tools	~3,105 tokens	−93%

The risk dividend: 21 of these 150 tools are critical-risk (destructive or financial) and cost 2,033 tokens (4% 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 Portkey Admin — 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 PORTKEY ADMIN TOKEN COST →

Instant setup, no code required.

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

//FAQ

Portkey Admin token-cost questions.

How many tokens does the Portkey Admin MCP server use?+

Its 150 tool definitions total 46,570 tokens — 23% 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 Portkey Admin 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 Portkey Admin's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Portkey Admin 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 931 tokens, a 98% 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 03-07-2026 from the PolicyLayer scan database over all 150 catalogued Portkey Admin tools. Counts refresh with every site build.