Skip to main content
Case StudyScore: 68 Silver

Why Slack Scores 68 for Agent Readiness (And What Keeps It From Gold)

Slack has one of the best developer APIs in the world. 200+ Web API methods. A real-time Events API. Block Kit for structured UI. 700+ OAuth scopes. And yet it scores 68 out of 100 — Silver tier, 7 points short of Gold. The reason is the same gap we see across every Silver-tier company: excellent infrastructure for human developers, but no agent-native discovery layer. Three files and a business decision separate Slack from Gold.

AH
AgentHermes Research
April 15, 202614 min read

Dimension-by-Dimension Breakdown

The Agent Readiness Score evaluates 9 dimensions, each weighted by importance to AI agents. Here is how Slack performs on every single one — what it gets right and where it falls short.

D1: Discovery

Partial
55
weight: 0.12

Slack has rich developer documentation and a sitemap, but no agent-card.json, no llms.txt, and no AGENTS.md. Agents discover Slack through brand recognition and search results — not through structured agent discovery protocols. The documentation is comprehensive but designed for human developers, not AI agent consumption.

D2: API Quality

Strong
85
weight: 0.15

Slack's strongest dimension. The Web API has 200+ methods with consistent JSON responses and clear error codes. The Events API provides real-time webhooks. Block Kit offers structured UI primitives. Every response is typed, documented, and versioned. This is enterprise-grade API design that agents can reliably interact with.

D3: Onboarding

Partial
62
weight: 0.08

App creation through api.slack.com/apps is self-service for standard tiers. Agents can obtain OAuth tokens via the standard flow. But workspace installation requires a human admin to approve. Enterprise Grid onboarding requires a sales process. An agent can create a Slack app but cannot get it installed in most workspaces autonomously.

D4: Pricing Transparency

Weak
38
weight: 0.05

Free and Pro plans are transparent with published pricing. But Business+ and Enterprise Grid — where most API-heavy usage happens — require "contact sales" for pricing. An AI procurement agent comparing collaboration tools cannot get Slack Enterprise pricing programmatically. This opacity drops D4 significantly.

D5: Payment Processing

Weak
32
weight: 0.08

No self-service enterprise billing API. An AI purchasing agent cannot upgrade a workspace to Business+ or Enterprise Grid via API. Standard tier purchases work through the Slack admin UI, but there is no programmatic checkout endpoint. Payment is human-driven at every tier above Pro.

D6: Data Quality

Strong
78
weight: 0.1

Consistent JSON response envelopes with an ok field, typed error codes, and pagination tokens. Block Kit provides structured message formatting. Metadata fields on messages enable programmatic categorization. Schema.org markup on the marketing site is limited but the API documentation itself is structured and thorough.

D7: Security

Strong
82
weight: 0.12

Excellent. OAuth 2.0 with granular scopes (700+ permission scopes). Token rotation support. Request signing for Events API webhooks (HMAC-SHA256). Audit log API for Enterprise Grid. security.txt published. SOC 2 Type II compliant. The only D7 gap: no x402 micropayment support, which is emerging tech.

D8: Reliability

Strong
80
weight: 0.13

status.slack.com exists with real-time incident tracking and historical uptime data. The API returns proper rate-limit headers (X-RateLimit-*) with Retry-After guidance. Tier-based rate limiting is well-documented by method. Health check endpoints exist for workspace connectivity testing.

D9: Agent Experience

Weak
48
weight: 0.1

This is where Slack loses the most ground. No agent-card.json at /.well-known/agent-card.json. No llms.txt at /llms.txt. No MCP server despite being one of the most MCP-connected services (community-built MCP servers exist, but Slack does not ship one officially). No structured way for agents to discover what Slack can do — they have to read human documentation.

68
overall score
85
D2 API Quality (best)
32
D5 Payment (worst)
7
points from Gold

Slack vs the Silver Tier

Slack sits in the middle of the Silver tier pack. What separates the one Gold company (Resend) from the rest is not API quality — it is agent-native features. Resend has agent-card.json and an MCP server. Nobody else in the top 6 does.

Company
Score
D9 Agent Exp
agent-card.json
MCP Server
Resend
75
72/100
Yes
Yes
Vercel
70
52/100
No
No
Supabase
69
50/100
No
No
Slack
68
48/100
No
No
Stripe
68
52/100
No
No
GitHub
67
46/100
No
No

The pattern is clear: Every company in the Silver tier has strong API fundamentals. What separates Silver from Gold is not better code — it is agent-native discovery. Resend cleared the bar with agent-card.json and an MCP server. Everyone else — Vercel, Supabase, Slack, Stripe, GitHub — is waiting to take the same step.

Four Changes That Push Slack to Gold

Slack needs 7 more points to reach Gold (75). Here are four changes, ordered by impact-to-effort ratio. The first two are afternoon projects. The third takes a couple of weeks. The fourth is a business decision, not a technical one.

Publish agent-card.json

/.well-known/agent-card.json

+3-4 points (D1 + D9)
30 minutes

Publish an A2A agent card declaring Slack's capabilities: messaging, channel management, user lookup, file sharing, workflow automation. This single file makes Slack discoverable by agent discovery protocols instead of requiring agents to know about Slack in advance.

Publish llms.txt

/llms.txt

+2-3 points (D1 + D9)
15 minutes

A plain-text summary of Slack's API surface optimized for LLM consumption. Which methods exist, what scopes they need, what Block Kit can do, how Events API works. Currently agents parse HTML documentation — llms.txt gives them a pre-digested summary.

Ship an official MCP server

npm package + registry listing

+4-5 points (D9 + agent-native bonus)
1-2 weeks

Community-built Slack MCP servers already exist. Slack should ship an official one with tools like send_message, create_channel, search_messages, list_users, and upload_file. This unlocks the agent-native bonus and makes Slack a first-class citizen in the MCP ecosystem.

Publish enterprise pricing via API

GET /api/pricing endpoint

+2-3 points (D4 + D5)
Business decision

Replace "contact sales" with structured pricing for Business+ and Enterprise Grid. Even a price range or per-seat band would help. AI procurement agents evaluating collaboration tools skip Slack Enterprise because they cannot get pricing without a human sales call.

The first two items alone (agent-card.json + llms.txt) would push Slack from 68 to an estimated 73-75 — right at the Gold threshold. Add an official MCP server and Slack reaches 77-78, the second-highest score in our 500-business dataset. The technical foundation is already world-class. Slack just needs to make it agent-discoverable.

This is the same pattern we documented in our Stripe breakdown and GitHub breakdown. The best developer platforms in the world are all 7-12 points from Gold because none of them have added the agent-native layer yet. The first one to do it wins the distinction that only Resend holds today.

Frequently Asked Questions

Why does Slack score the same as Stripe (68) despite being very different products?

Both have excellent APIs (Slack D2: 85, Stripe D2: 82) and strong security (Slack D7: 82, Stripe D7: 78). Both lose points on agent-native features (no agent-card.json, no official MCP server). And both have pricing opacity — Stripe for enterprise Connect pricing, Slack for Business+ and Enterprise Grid. The Agent Readiness Score measures how easily an AI agent can discover, understand, and use a service. Both Slack and Stripe are built for developers but not yet optimized for autonomous agent interaction.

Slack has community-built MCP servers. Why does that not count?

Community MCP servers are not discoverable at Slack's domain. An agent checking slack.com for MCP capabilities finds nothing. The community server on GitHub requires manual discovery, manual installation, and manual configuration. For agent readiness, what matters is what an agent finds when it visits your domain and checks standard paths — /.well-known/agent-card.json, /llms.txt, and MCP endpoints linked from those files.

Could Slack realistically reach Gold (75+)?

Yes, with relatively modest effort. Slack starts at 68 — only 7 points from Gold. agent-card.json (+3-4 points) and llms.txt (+2-3 points) together could push it past 75 in an afternoon. An official MCP server (+4-5 points) would push it to 77-78, solidly Gold. The technical foundation is already there — Slack just needs to add the agent-native discovery layer on top.

How does Slack compare to Microsoft Teams for agent readiness?

Microsoft Teams has a larger API surface through Microsoft Graph but scores lower on agent readiness overall (estimated 58-62 Silver). Teams requires Azure AD authentication which is more complex for agents to negotiate. Teams has no agent-card.json, no llms.txt, and no official MCP server either. Slack's advantage is simpler OAuth scopes and more consistent API design — agents make fewer errors interacting with Slack's API than with Microsoft Graph.


See how your company compares to Slack

Run a free Agent Readiness Scan and get your dimension-by-dimension breakdown. See your score, your tier, and exactly what to fix — in 60 seconds.


Share this article: