// MCP TOKEN COST

The Meridian MCP server costs 104,466 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 Meridian MCP server's 235 tool definitions consume 104,466 tokens — 52% of a 200k context window, and 55× the median MCP server (1,889 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS tiktoken o200k_base · rank #12 of 9,565 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 52%

1M WINDOW 10%

Corpus context: Meridian ranks #12 of 9,565 measured MCP servers by definition cost. The median is 1,889 tokens, p90 is 13,337, and the heaviest (Ainumbers Mcp Apps) is 345,766 — 173% 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 104,466 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 445 tokens per tool.

Tool	Category	Tokens	% of server
generate_handoff	Write	3,404	3.3%
update_sprint_item	Write	2,689	2.6%
add_sprint_item	Write	2,350	2.2%
add_sprint_item_pointer	Write	2,161	2.1%
complete_sprint_item	Write	1,584	1.5%
accept_handoff	Destructive	1,508	1.4%
update_paragraph	Write	1,241	1.2%
execute_batch	Execute	1,184	1.1%
claim_sprint_item	Destructive	1,066	1.0%
record_handoff_correction	Destructive	1,053	1.0%
start_session	Execute	1,031	1.0%
fan_out_sprint_items	Destructive	1,025	1.0%
commit_proposal_promotion	Write	939	0.9%
batch_read	Read	922	0.9%
reconcile_stale_claims	Destructive	903	0.9%
batch_mutate	Destructive	900	0.9%
configure_wave_gate	Write	886	0.8%
prospect_symbol	Execute	787	0.8%
transfer_sprint_item_claim	Financial	787	0.8%
save_profile_layer	Write	767	0.7%
update_workspace_settings	Write	766	0.7%
add_custom_hook	Write	761	0.7%
set_executor_config	Write	760	0.7%
request_hitl	Destructive	734	0.7%
relocate_sprint_item_pointer	Destructive	731	0.7%
insert_equation	Write	731	0.7%
complete_wave_gate	Write	722	0.7%
verify_handoff_token	Read	716	0.7%
set_capability_profile	Write	701	0.7%
link_flag_to_section	Destructive	694	0.7%
start_research_run	Execute	690	0.7%
set_ai_log_export_config	Write	687	0.7%
resume_wave	Write	679	0.6%
register_session_recovery	Write	670	0.6%
claim_file	Destructive	654	0.6%
update_custom_hook	Write	649	0.6%
add_proposal_gate	Write	639	0.6%
search_outputs	Read	638	0.6%
register_docx_derivative	Write	634	0.6%
add_note	Write	630	0.6%
index_table	Destructive	624	0.6%
start_wave_run	Execute	623	0.6%
release_sprint_item_claim	Destructive	604	0.6%
ingest_document	Destructive	601	0.6%
add_proposal	Write	601	0.6%
annotate_outputs	Write	593	0.6%
create_proposal_successor	Write	592	0.6%
complete_experiment_run	Write	584	0.6%
index_figure	Destructive	583	0.6%
index_equation	Destructive	570	0.5%
complete_research_run	Write	566	0.5%
promote_proposal	Write	550	0.5%
link_figure_caption	Destructive	533	0.5%
finalize_wave_run	Financial	531	0.5%
preview_proposal_promotion	Read	529	0.5%
link_table_caption	Destructive	527	0.5%
get_workspace_proposals	Read	527	0.5%
export_ai_log_otel	Write	525	0.5%
ingest_document_structure	Destructive	520	0.5%
resolve_sprint_item_pointers	Write	519	0.5%
search_code_semantic	Read	511	0.5%
set_capability_manifest	Write	506	0.5%
get_effective_capability_profile	Read	502	0.5%
save_watchlist_query	Write	502	0.5%
register_external_job	Write	500	0.5%
resolve_citations	Write	500	0.5%
set_active_repo	Write	495	0.5%
verify_docx_diff	Read	494	0.5%
start_remote_task	Execute	490	0.5%
start_experiment_run	Execute	483	0.5%
assign_sprint_waves	Write	482	0.5%
update_md_section	Write	478	0.5%
merge_project	Write	469	0.4%
promote_docx_candidate	Write	468	0.4%
link_proposal_lineage	Destructive	465	0.4%
clone_profile_layer	Destructive	462	0.4%
get_sprint_items	Read	462	0.4%
analyze_model_efficiency	Read	458	0.4%
search_ai_log	Read	458	0.4%
acquire_docx_document_lease	Destructive	455	0.4%
link_manual_github_issue	Write	452	0.4%
get_flag_drift	Read	450	0.4%
claim_docx_region	Destructive	449	0.4%
paper_search	Read	449	0.4%
resolve_proposal_gate	Write	447	0.4%
save_finding	Write	442	0.4%
add_subtask	Write	441	0.4%
find_similar_figure	Read	440	0.4%
get_notes	Read	440	0.4%
purge_ai_log	Destructive	438	0.4%
get_latex_structure	Read	435	0.4%
request_manual_issue_screening_toggle	Write	426	0.4%
set_parent_project	Write	422	0.4%
set_project_execution_mode	Write	418	0.4%
update_decision	Write	418	0.4%
activate_profile_layer	Write	416	0.4%
record_experiment_event	Destructive	413	0.4%
checkpoint	Destructive	411	0.4%
run_watchlist_query	Execute	409	0.4%
get_tunnel_diagnostics	Read	409	0.4%
validate_assumption	Read	404	0.4%
pin_decision	Write	400	0.4%
save_blog_post	Write	399	0.4%
run_verification	Execute	398	0.4%
find_symbol_usages	Read	395	0.4%
update_external_job	Write	395	0.4%
add_insight	Write	394	0.4%
get_parallelizable_groups	Read	393	0.4%
reset_plugin_override	Destructive	392	0.4%
store_finding	Destructive	388	0.4%
get_server_logs	Read	388	0.4%
find_outputs_by_source	Read	385	0.4%
search_server_logs	Read	385	0.4%
get_effective_profile	Read	382	0.4%
move_workspace_note_to_project	Write	380	0.4%
add_sprint_note	Write	379	0.4%
capture_research_finding	Destructive	376	0.4%
find_orphaned_docx_staged_files	Read	370	0.4%
register_run_artifact	Write	370	0.4%
analyze_sprint	Read	364	0.3%
get_remote_task_status	Read	353	0.3%
reset_profile_layer	Destructive	351	0.3%
send_message	Write	351	0.3%
create_experiment	Write	348	0.3%
update_workspace_sprint_item	Write	348	0.3%
add_workspace_sprint_item	Write	347	0.3%
delete_sprint_item_pointer	Destructive	344	0.3%
delete_custom_hook	Destructive	343	0.3%
reopen_proposal_gate	Destructive	343	0.3%
log_task	Destructive	341	0.3%
get_citation_edges	Read	341	0.3%
complete_external_job	Write	340	0.3%
add_workspace_proposal	Write	339	0.3%
promote_research_run	Write	339	0.3%
search_all	Read	338	0.3%
get_capability_manifest	Read	335	0.3%
register_session	Write	329	0.3%
set_sprint	Write	327	0.3%
refresh_context	Destructive	326	0.3%
split_sprint_item	Destructive	325	0.3%
create_project	Write	325	0.3%
set_agent_instructions	Write	324	0.3%
get_document_structure	Read	320	0.3%
merge_sprint_items	Write	319	0.3%
promote_experiment_run	Write	319	0.3%
clear_capability_profile	Destructive	318	0.3%
release_docx_region_claims	Destructive	315	0.3%
get_findings	Read	313	0.3%
advance_proposal_status	Read	309	0.3%
get_server_log_checkpoint	Read	306	0.3%
idle_until_all_done	Destructive	301	0.3%
github_search	Read	295	0.3%
delete_watchlist_query	Destructive	294	0.3%
receive_messages	Destructive	293	0.3%
snapshot_graph_metrics	Read	287	0.3%
set_north_star	Write	285	0.3%
get_workspace_sprint_items	Read	284	0.3%
rename_project	Write	284	0.3%
release_docx_document_lease	Destructive	282	0.3%
add_workspace_note	Write	282	0.3%
get_planning_brief	Read	279	0.3%
pin_workspace_decision	Write	279	0.3%
find_similar_equation	Read	277	0.3%
get_docx_document_lease	Read	276	0.3%
get_flag_registry	Read	273	0.3%
export_ai_log	Write	273	0.3%
export_ai_log_artifacts	Write	270	0.3%
get_session_brief	Read	269	0.3%
find_similar_table	Read	267	0.3%
answer_hitl	Destructive	265	0.3%
get_docx_region_claims	Read	264	0.3%
get_ai_log_export_status	Read	261	0.2%
list_worktrees_pending_cleanup	Read	260	0.2%
list_hitl_requests	Read	258	0.2%
set_goal	Write	256	0.2%
archive_decision	Write	255	0.2%
get_sprint_progress	Read	247	0.2%
dismiss_hitl	Destructive	239	0.2%
release_file	Destructive	233	0.2%
social_search	Read	233	0.2%
get_pinned_decisions	Read	231	0.2%
complete_workspace_sprint_item	Write	231	0.2%
get_proposal_gates	Read	224	0.2%
list_profile_layers	Read	224	0.2%
reconcile_sprint_drift	Read	224	0.2%
get_profile_layer	Read	219	0.2%
delete_note	Destructive	217	0.2%
get_proposal_lineage	Read	214	0.2%
get_context_block	Read	209	0.2%
refresh_tool_manifest	Read	209	0.2%
get_connection_log	Read	208	0.2%
get_file_claims	Read	199	0.2%
compare_proposal_versions	Read	197	0.2%
idle_until_session_done	Read	197	0.2%
list_active_worktrees	Read	196	0.2%
load_handoff	Read	194	0.2%
search_synthesis	Read	181	0.2%
get_session_recovery	Read	179	0.2%
get_sprint_item_pointers	Read	173	0.2%
get_custom_hooks	Read	168	0.2%
list_remote_tasks	Read	166	0.2%
list_research_runs	Read	155	0.1%
list_external_jobs	Read	153	0.1%
read_note	Read	151	0.1%
get_profile_layer_revisions	Read	150	0.1%
list_watchlist_queries	Read	149	0.1%
get_experiment_events	Read	146	0.1%
list_resumable_sessions	Read	146	0.1%
get_session_log	Read	142	0.1%
list_experiment_runs	Read	138	0.1%
get_insights	Read	134	0.1%
get_session_activity	Read	133	0.1%
list_sessions	Read	120	0.1%
get_external_job	Read	118	0.1%
get_goal	Read	118	0.1%
get_plugin_details	Read	116	0.1%
list_experiments	Read	113	0.1%
search_tasks	Read	113	0.1%
get_agent_instructions	Read	110	0.1%
get_sprint_notes	Read	110	0.1%
get_graph_diff	Read	108	0.1%
list_plugins	Read	100	0.1%
get_experiment_run	Read	97	0.1%
get_experiment	Read	94	0.1%
get_research_run	Read	94	0.1%
get_project_by_name	Read	92	0.1%
get_blog_posts	Read	83	0.1%
get_tasks	Read	83	0.1%
list_projects	Read	82	0.1%
get_symbol_hotspots	Read	81	0.1%
get_hitl_request	Read	70	0.1%
get_workspace_settings	Read	64	0.1%
get_symbol_claims	Read	61	0.1%
get_workspace_decisions	Read	56	0.1%
get_workspace_notes	Read	52	0.0%

// CUT THE COST

Your agent uses a handful of these tools. It pays for all 235.

You don't need all 235 of those definitions in the window. PolicyLayer is an MCP gateway that sits in front of Meridian: 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 235 tools (no gateway)	104,466 tokens	—
3 granted tools	~1,334 tokens	−99%
5 granted tools	~2,223 tokens	−98%
10 granted tools	~4,445 tokens	−96%

The risk dividend: 47 of these 235 tools are critical-risk (destructive or financial) and cost 24,527 tokens (23% 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 Meridian — 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 MERIDIAN TOKEN COST →

Instant setup, no code required.

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

//FAQ

Meridian token-cost questions.

How many tokens does the Meridian MCP server use?+

Its 235 tool definitions total 104,466 tokens — 52% 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 Meridian 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 Meridian's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Meridian 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 1,334 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 02-10-2026 from the PolicyLayer scan database over all 235 catalogued Meridian tools. Counts refresh with every site build.