Every conversation with the Slack MCP server begins with 8 tool definitions already in context — about 681 tokens of it. Against the 19 static rules in the catalogue it scores 99 out of 100 — grade A — raising 1 scored findings in 1 of the six categories.
The spread runs from 100/100 for surface down to 96/100 for design. It is catalogued here under Communication, alongside other servers agents reach for in the same kind of work.
Measured against a 200,000-token window, the surface consumes about 0% of it before the conversation starts, and keeps consuming it on every turn. That averages 85 tokens per tool definition, counted with the o200k_base encoding.
That lands exactly on the Communication median of 681 tokens.
Every finding below names the rule behind it and the tool it fired on.
`design/overlap-cluster` accounts for one finding. Tools whose descriptions substantially overlap force the model to pick between near-synonyms — the classic cause of wrong-tool calls. Clusters are usually one tool wearing several names; consolidate with include-flags or a mode parameter. It applies to the surface as a whole rather than to one tool.
Findings that measurably degrade tool selection.
Tools whose descriptions substantially overlap force the model to pick between near-synonyms — the classic cause of wrong-tool calls. Clusters are usually one tool wearing several names; consolidate with include-flags or a mode parameter.
slack_get_channel_history, slack_get_thread_replies, slack_get_users, slack_get_user_profile
Slack scores a full 100 in surface, naming, descriptions, schemas and annotations, which is to say the surface stays inside a sensible tool and token budget, the tool names follow one convention end to end, every tool carries a description long enough to disambiguate it, the input schemas are typed tightly enough to constrain what a model sends and read and write operations are machine-distinguishable before a client runs them.
`slack_reply_to_thread` is the most expensive definition at 124 tokens — about 18% of the whole surface — against 51 for `slack_get_user_profile`. Every tool carries a description, ranging from 33 to 77 characters.
None declare annotations, so a client cannot tell a read from a write without reading the prose. Across the surface, schemas carry 2 top-level parameters on average and reach a maximum depth of 1.
All 8 tools as published, heaviest first. Token counts use the o200k_base encoding and are approximate.
| Tool | Tokens | Description | Params | Required | Depth | Annotations |
|---|---|---|---|---|---|---|
slack_reply_to_thread | 124 | 43 chars | 3 | 3 | 1 | no |
slack_get_thread_replies | 110 | 35 chars | 2 | 2 | 1 | no |
slack_add_reaction | 95 | 33 chars | 3 | 3 | 1 | no |
slack_get_users | 80 | 77 chars | 2 | 0 | 1 | no |
slack_list_channels | 76 | 68 chars | 2 | 0 | 1 | no |
slack_get_channel_history | 73 | 34 chars | 2 | 1 | 1 | no |
slack_post_message | 70 | 37 chars | 2 | 2 | 1 | no |
slack_get_user_profile | 51 | 52 chars | 1 | 1 | 1 | no |
Within Communication this is rank 1 of 1 scanned servers — exactly on the category median of 99.
The closest comparisons in this directory are Cloudflare Docs, Context7 and PostgreSQL, which sit nearest on score and face the same kind of surface-design pressure.
A fix plan turns these findings into an ordered set of changes with the token saving of each one.
This page is a static analysis of Slack's published tools/list surface on the date shown. It is not a security review, a performance benchmark, or a judgement of the product behind the surface. Read how the score is computed, or browse every audited server.