For Claude users
Your reps already live in Claude. Clinch is how Claude knows your deals.
Native Claude.ai connector with read and write. Generate pre-call briefs inside Claude, log post-call outcomes by saying one sentence, ask Claude how you are doing against any competitor and get your own numbers back. Five first-party skills plus a 25-tool MCP server. Pro tier.
What is MCP
The integration layer for AI-native work.
The Model Context Protocol, created by Anthropic in November 2024 and donated to the Linux Foundation’s Agentic AI Foundation in December 2025, is an open protocol that lets AI assistants query external tools and data sources with user permission.
The practical result: instead of copy-pasting between your CI tool and Claude, Claude calls the CI tool directly. Your data stays in Clinch. Your workflow stays in Claude. The two become one surface.
Why this matters for CI
Your buyer already works inside Claude. Meet them there.
Competitive intelligence is a research discipline. Most PMMs, founders, and sales leaders already do their research inside Claude or ChatGPT. Asking them to switch to a dashboard every time they want to understand a competitor is the old pattern. MCP collapses the switch.
No context switch
Claude queries Clinch while you keep working. No tab, no login, no dashboard navigation.
Real data, not search results
When you ask Claude about a competitor, it answers from your Clinch timeline, not a web search.
Your ongoing AI gets smarter
Every battlecard, every snapshot, every pricing change becomes an input Claude can reason over.
First-party Claude skills
Skills meet the rep where the work happens.
Each skill is one workflow. They come with the Clinch plugin, so adding Clinch from the Claude directory installs all of them. Trigger them by speaking to Claude in plain English. They compose with the Calendar, Krisp, and Zoom connectors the rep may already have.
Skill 1
Pre-call brief
“I have a call in 20 minutes with Acme, evaluating us vs Klue.”
Claude reads the calendar event, resolves the prospect domain, calls generate_deal_brief, and drops a one-pager into the chat. The rep reads it in 60 seconds.
Skill 2
Post-call capture
“We lost that to Klue, integration concerns.”
Claude extracts the competitor, the outcome, and the reason in the rep's own words, then writes it back via log_deal_outcome. The dashboard fills itself.
Skill 3
Deal coach
“How am I doing against Klue this quarter?”
Claude pulls the rep's personal win/loss record via get_my_competitive_record, names the recurring loss reason, and offers to coach with the live battlecard.
Setup
Five steps. About 3 minutes to your first answer.
- 01
Start Clinch Pro or a free trial
Clinch for Claude is included on Clinch Pro at $149 per month and in the 14 day trial. It unlocks as soon as your workspace is on either.
- 02
Add Clinch from the Claude directory
In Claude, open Customize, then Plugins, then Discover, search for Clinch, and click Add. The plugin brings the Clinch connector, the guided setup and five skills, and it updates itself. The Claude page in your Clinch dashboard links straight to it.
- 03
Connect and approve
Claude asks you to connect Clinch. Sign in and click Allow on the consent screen (OAuth 2.0 authorization code with PKCE). Access tokens last 1 hour and Claude refreshes them automatically.
- 04
Run the guided setup
In a new chat, type /clinch:setup. Claude reads the sources you choose, such as your CRM, docs and call recordings, proposes competitors and rivalry notes, and saves nothing until you approve it.
- 05
Ask Claude anything and log activity
All 25 tools are live (12 read, 13 write). The skills trigger on plain phrases, so reps never have to remember tool names. On Claude Team or Enterprise, an owner can turn the plugin on for the whole organization. Teams that cannot use plugins can add the Clinch connector on its own from the directory.
What to ask Claude
Real queries, real answers.
These are the actual prompts teams run against the Clinch connector. Pick the role that fits you.
What a PMM can ask
The PMM uses Claude as an always-on analyst, with the full Clinch timeline, battlecards, and win/loss data underneath.
“Summarize every pricing change across my tracked competitors in the last 90 days.”
Claude queries the Clinch pricing_history tool, groups by competitor, and flags the two that changed twice.
“Which competitor has been most active this quarter and what are the three biggest moves?”
Claude uses competitor_timeline plus search_competitive_changes to rank by change frequency and summarize the top moves.
“Draft a one-page competitive brief for my board meeting next Tuesday.”
Claude pulls the latest snapshots, battlecard summaries, and trend rows, then writes the brief in your voice.
What a founder can ask
Without a PMM hire, the founder uses Claude as the CI function. Every question is answered against real data, not a web search.
“What changed with Acme Corp since our last fundraise?”
Claude queries competitor_timeline with a date window and returns a ranked change list with context.
“Is there a pattern across my three closest rivals I should worry about?”
Claude uses competitive_landscape to surface parallel moves (e.g., three pricing shifts in the same window).
“What are the top three competitive objections I should prep for in next week's fundraise?”
Claude reads your battlecards plus your positioning template and writes objection responses a VC would accept.
What a VP of Sales can ask
Competitive enablement and pipeline analysis, answered in natural language against the same data the battlecards are built on.
“Which competitor has the most deal pressure on our pipeline right now?”
Claude combines deal_outcomes with search_competitive_changes to identify the rival showing up most in recent losses.
“Give me the five-minute prep brief for tomorrow's call with Globex.”
Claude fetches the Clinch battlecard, adds recent change context, and shapes it into a verbal briefing.
“How has our win rate against Acme moved quarter over quarter, and what changed on their side?”
Claude correlates deal_outcomes with competitor_timeline to explain the trendline, not just report it.
What an AE can ask
Live-call prep and post-call debrief, answered on a phone between meetings without opening a dashboard.
“I have a call in 10 minutes with a prospect evaluating Klue. Give me a landmine question and two objections to prep for.”
Claude fetches the Klue battlecard, returns a specific landmine question and the top two objections with responses.
“What changed with Crayon in the last 30 days that might come up in a call?”
Claude uses competitor_timeline scoped to 30 days and returns bullets the AE can skim in 20 seconds.
What a sales enablement lead can ask
Program design, adoption measurement, and rep coaching, each grounded in the actual battlecard and deal data.
“Which battlecards have been viewed the fewest times this month, and which competitors are in the most deals?”
Claude surfaces the mismatch between adoption and market pressure so you know where to focus rep coaching.
“Draft a 15-minute rep certification quiz on our top three competitors.”
Claude reads your battlecards and generates scenario-based questions a new AE must answer to pass certification.
For builders
Technical details.
Endpoint
Streamable HTTP MCP server at /api/mcp. Manual JSON-RPC over standard HTTP. No SSE.
Authentication
OAuth 2.0 authorization code + PKCE (RFC 6749, RFC 7636). Dynamic client registration via RFC 7591.
Token lifetime
Access tokens: 1 hour. Refresh tokens: 90 days, rotated on use. Revocable from Clinch settings.
Rate limits
60 requests per minute per access token, counted in Postgres. If the count cannot be read, the request is refused rather than let through.
Scopes
Two scopes: read (twelve lookup tools) and write (thirteen mutation tools). Connector requests both by default; user approves each in the consent screen. Tokens carry the granted scopes for their full lifetime.
Twenty-five tools
Read: get_competitive_landscape, get_competitor_timeline, get_pricing_history, get_battlecard, search_competitive_changes, get_latest_digest, get_saved_briefs, get_team_scoreboard, get_my_competitive_record, get_company_profile, get_competitive_context, list_unconfirmed_items. Write: complete_setup, log_deal_outcome, log_competitor_mention, flag_battlecard_section, request_brief, generate_deal_brief, add_competitor, update_competitor, set_competitive_context, replace_competitive_context, set_company_profile_field, replace_company_profile_field, review_logged_item. Every write lands in an audit log.
Discovery
RFC 8414 metadata at /.well-known/oauth-authorization-server. Protected resource metadata at /.well-known/oauth-protected-resource.
Frequently asked
Questions worth answering.
- What is the Model Context Protocol (MCP)?
- MCP is an open protocol that lets AI assistants like Claude, ChatGPT, and Copilot query external tools and data sources directly with user permission. Created by Anthropic in November 2024 and donated to the Linux Foundation's Agentic AI Foundation in December 2025, MCP is the de facto integration layer for AI-native workflows as of 2026.
- Why does MCP matter for competitive intelligence?
- CI is a research discipline, and most buyers already do their CI research inside Claude, ChatGPT, or Perplexity. MCP turns your CI tool into a native part of that workflow. Instead of switching tabs to a dashboard, you ask Claude directly and Claude answers using your real Clinch data.
- Is the MCP connector included or is it an add-on?
- Included on Clinch Pro ($149 per month). Not included in the Clinch plan. There is no separate add-on fee. MCP API usage does not count against any additional quota.
- What tools does the Clinch MCP server expose?
- Twenty-five tools split across two scopes. Twelve read tools: get_competitive_landscape, get_competitor_timeline, get_pricing_history, get_battlecard, search_competitive_changes, get_latest_digest, get_saved_briefs, get_team_scoreboard, get_my_competitive_record, get_company_profile, get_competitive_context, list_unconfirmed_items. Thirteen write tools (under a separate write scope): complete_setup, log_deal_outcome, log_competitor_mention, flag_battlecard_section, request_brief, generate_deal_brief, add_competitor, update_competitor, set_competitive_context, replace_competitive_context, set_company_profile_field, replace_company_profile_field, review_logged_item. Every write is attributed to the calling user and audited, and the tools that can overwrite or delete (update_competitor, the two replace tools, review_logged_item) always ask you before they run.
- What about the Clinch skills?
- Skills are workflow playbooks that come with the Clinch plugin, so adding the plugin from the Claude directory installs all of them and keeps them current. Five first-party skills today: clinch-pre-call-brief (triggers when you have a competitive call upcoming, generates the brief), clinch-post-call-capture (triggers after a call to log outcomes back to Clinch), clinch-deal-coach (triggers on questions about your own win rate or live objection handlers), clinch-import-from-crm (brings closed deals from your CRM connector into Clinch), and clinch-onboard-workspace (sets up a new workspace from your CRM, docs, and transcripts). Each skill calls the MCP tools above; the connector works fully without them.
- Is Claude's write access safe?
- Write tools require an explicit write scope you approve on the consent screen. The everyday tools only add: they log deal outcomes and mentions, request briefs, add competitors, and fill fields that are empty, so they never overwrite your company profile or rivalry notes. The four tools that can overwrite or delete (update_competitor, replace_competitive_context, replace_company_profile_field, review_logged_item) make Claude ask you before each call. Every write lands in an audit log scoped to your org, and idempotency keys make retries safe.
- Does the MCP server work with ChatGPT or only Claude?
- As of April 2026 the MCP ecosystem is strongest inside Claude.ai and desktop Claude. OpenAI announced MCP support in April 2025. Microsoft adopted in July 2025. Practically, Claude.ai has the most mature connector UX today. We test and document against Claude.ai first; support for other MCP-compatible clients follows as those clients mature.
- How does authentication work?
- OAuth 2.0 authorization code flow with PKCE (RFC 6749 + RFC 7636). Claude.ai registers as an MCP client via dynamic client registration (RFC 7591). Access tokens live one hour; refresh tokens live 90 days and rotate on use. You can revoke access any time from Clinch settings; revocation is immediate across all Claude sessions.
- Is my competitive data safe?
- Yes. The MCP server is org-scoped, and write access is a separate scope you approve on the consent screen. Every request is bearer-token authenticated and validated against your org. No other customer can access your data through the connector. Tokens are storage-encrypted and never logged. Rate limits (60 requests per minute per token) protect against runaway agents.
- Do I have to use MCP to use Clinch?
- No. The core Clinch workflow (snapshots, digest, battlecards) works without Claude. MCP is the Pro-tier enhancement for teams that already live inside Claude.ai. If you do not use Claude, skip MCP and stay on the Clinch plan.
Need help with Clinch in Claude?
Email hello@getclinch.ai. If you have a Clinch account, you can also open a ticket from Help in your workspace. Tell us what you asked Claude and what happened, and we will reply by email.