// MCP TOKEN COST

The Dialogbrain MCP server costs 38,562 tokens before the first call.

Connect Dialogbrain and its 157 tool definitions are loaded into the model's context on every request — 19% of a 200k window spent before your agent does anything.

QUICK ANSWER The Dialogbrain MCP server's tool definitions consume 38,562 tokens — 20× the median MCP server (1,905 tokens). A scoped grant exposing only the tools you use cuts that roughly in proportion.

MEASURED FROM SCHEMAS 157 tools · 38,562 tokens · 19% of 200k · 3.9% 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 19%

1M WINDOW 3.9%

Corpus context: Dialogbrain ranks #16 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 38,562 tokens go.

Each row is one tool definition as a tools/list entry — name, description and input schema — counted with o200k_base. Average: 246 tokens per tool.

Tool	Category	Tokens	% of server
agents_update	Write	3,142	8.1%
documents_create	Write	2,543	6.6%
widgets_update	Write	1,001	2.6%
videos_generate	Write	997	2.6%
agents_trigger_create	Execute	885	2.3%
agents_simulate_inbound	Execute	708	1.8%
calls_send_to_telegram_call	Write	689	1.8%
calls_make	Write	529	1.4%
calls_send_to_meet	Write	496	1.3%
calls_list_history	Read	490	1.3%
agents_traces_list	Read	485	1.3%
reminder_set	Write	462	1.2%
messages_send	Write	456	1.2%
knowledge_query	Read	448	1.2%
ai_filters_create	Write	422	1.1%
widgets_create	Write	419	1.1%
agent_handoff	Write	417	1.1%
messages_read_history	Read	414	1.1%
notes_save	Write	408	1.1%
calendar_create_event	Write	406	1.1%
feedback_save	Write	392	1.0%
agents_trigger_update	Execute	387	1.0%
linkedin_update_profile	Write	368	1.0%
contacts_update	Write	364	0.9%
ai_tags_create	Write	356	0.9%
kg_get_relationships	Read	355	0.9%
web_fetch	Read	343	0.9%
youtube_upload_video	Write	340	0.9%
agents_activity	Execute	339	0.9%
contacts_find	Read	335	0.9%
agents_create	Write	333	0.9%
linkedin_search	Read	303	0.8%
ai_filters_update	Write	289	0.7%
agents_ask	Read	286	0.7%
kg_find_entity	Read	283	0.7%
files_read	Read	274	0.7%
search_links	Read	273	0.7%
ai_tags_update	Write	270	0.7%
agents_trace_get	Read	269	0.7%
search_threads	Read	268	0.7%
group_list	Read	265	0.7%
notes_recall	Read	255	0.7%
images_generate	Write	251	0.7%
search_files	Read	246	0.6%
images_search	Read	245	0.6%
notes_search	Read	243	0.6%
group_create	Write	242	0.6%
tasks_create	Write	238	0.6%
group_promote_admin	Write	228	0.6%
calendar_check_availability	Read	226	0.6%
ai_filters_test	Read	220	0.6%
threads_update	Write	218	0.6%
messages_send_email	Write	216	0.6%
messages_forward	Write	214	0.6%
group_search	Read	213	0.6%
calendar_update_event	Write	213	0.6%
calls_wait	Execute	211	0.5%
files_ingest	Write	206	0.5%
prompts_update	Write	203	0.5%
files_upload	Write	201	0.5%
web__local_search	Read	200	0.5%
ai_filters_list	Read	198	0.5%
search_messages	Read	195	0.5%
instagram_publish_media	Write	195	0.5%
contacts_add_channel	Write	192	0.5%
youtube_video_query	Read	191	0.5%
tasks_list	Read	187	0.5%
linkedin_search_filters	Read	185	0.5%
youtube_update_video	Write	183	0.5%
calendar_list_events	Read	181	0.5%
agents_traces_stats	Read	180	0.5%
group_add	Write	178	0.5%
web_search	Read	175	0.5%
group_add_member	Write	173	0.4%
group_join	Execute	170	0.4%
tasks_update	Write	166	0.4%
workspace_switch	Write	166	0.4%
agents_add_file	Write	164	0.4%
agents_list	Read	162	0.4%
agents_list_drafts	Read	160	0.4%
agents_update_from_template	Write	155	0.4%
agents_approve_draft	Write	154	0.4%
linkedin_raw_request	Write	153	0.4%
ai_tags_add_to_thread	Write	152	0.4%
prompts_prompt_restore	Write	151	0.4%
agents_prompt_history	Write	150	0.4%
youtube_post_comment_reply	Write	150	0.4%
group_preview_messages	Read	149	0.4%
agent_silence	Write	148	0.4%
youtube_list_comments	Read	147	0.4%
ai_tags_list	Read	144	0.4%
files_get_base64	Read	143	0.4%
prompts_prompt_history	Write	142	0.4%
calls_get_transcript	Read	140	0.4%
agents_reject_draft	Write	140	0.4%
agents_prompt_restore	Write	139	0.4%
present_tab	Write	137	0.4%
channels_connect_telegram_bot	Write	135	0.4%
contacts_discover	Write	135	0.4%
calls_hangup	Write	132	0.3%
group_scan	Read	131	0.3%
folders_create	Write	126	0.3%
calls_list_active	Read	125	0.3%
linkedin_invite	Write	118	0.3%
ai_tags_remove_from_thread	Destructive	116	0.3%
agents_list_files	Read	116	0.3%
contacts_profile	Read	116	0.3%
system_sleep	Execute	114	0.3%
files_info	Read	113	0.3%
reminder_list	Read	112	0.3%
agents_remove_file	Destructive	109	0.3%
vision_query	Read	109	0.3%
calendar_delete_event	Destructive	107	0.3%
collections_delete	Destructive	107	0.3%
agents_get	Read	107	0.3%
folders_delete	Destructive	105	0.3%
notes_delete	Destructive	103	0.3%
youtube_moderate_comment	Write	103	0.3%
collections_add_file	Write	102	0.3%
youtube_list_videos	Read	101	0.3%
widgets_get_embed_code	Read	100	0.3%
collections_list_files	Read	99	0.3%
agents_task_complete	Write	99	0.3%
collections_create	Write	98	0.3%
ai_filters_delete	Destructive	97	0.3%
collections_assign_agent	Write	92	0.2%
ai_tags_delete	Destructive	91	0.2%
instagram_list_media	Read	91	0.2%
job_update_context	Write	91	0.2%
collections_unassign_agent	Destructive	89	0.2%
messages_delete	Destructive	89	0.2%
instagram_update_media	Write	89	0.2%
contacts_sync	Write	87	0.2%
job_escalate	Write	85	0.2%
collections_list	Read	84	0.2%
linkedin_list_reactions	Read	84	0.2%
job_complete	Write	84	0.2%
collections_remove_file	Destructive	83	0.2%
widgets_delete	Destructive	83	0.2%
linkedin_add_comment	Write	83	0.2%
youtube_delete_video	Destructive	82	0.2%
widgets_get	Read	82	0.2%
linkedin_get_profile	Read	81	0.2%
agents_trigger_delete	Destructive	80	0.2%
widgets_list	Read	78	0.2%
reminder_cancel	Destructive	76	0.2%
tasks_delete	Destructive	76	0.2%
youtube_delete_comment	Destructive	76	0.2%
linkedin_list_invitations_sent	Read	75	0.2%
linkedin_list_connections	Read	74	0.2%
prompts_get	Read	74	0.2%
agents_delete	Destructive	67	0.2%
job_read_context	Read	66	0.2%
linkedin_get_company	Read	65	0.2%
workspace_list	Read	64	0.2%
prompts_list	Read	57	0.1%
workspace_current	Write	56	0.1%

// SCOPED GRANT

Most agents use a handful of these tools. They pay for all 157.

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 (246 tokens each).

Grant scope	Definition cost	Reduction
All 157 tools (no gateway)	38,562 tokens	—
3 granted tools	~737 tokens	−98%
5 granted tools	~1,228 tokens	−97%
10 granted tools	~2,456 tokens	−94%

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

//FAQ

Dialogbrain token-cost questions.

How many tokens does the Dialogbrain MCP server use?+

Its 157 tool definitions total 38,562 tokens — 19% 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 Dialogbrain 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 Dialogbrain's token usage?+

Expose fewer tools. A PolicyLayer grant scopes Dialogbrain 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 737 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 07-06-2026 from the PolicyLayer scan database over all 157 catalogued Dialogbrain tools. Counts refresh with every site build.