MCP Tools

The Caesura MCP server currently exposes the following tools for use by your connected AI agents.

warning

These tools are under active development. Schemas and behaviors may evolve.

---

analyze_transcript

Analyzes a conversation snippet for psychological patterns, communication dynamics, emotional shifts, and actionable next-step recommendations.

The tool is stateless—you must include enough context in each call for a meaningful analysis. Overlapping conversation chunks between subsequent calls are fine and expected.

Inputs

Parameter	Type	Required	Description
transcript	string	Yes	The conversation fragment to analyze. Format as one line per turn (e.g., Speaker: text). Prior turns can be summarized in parentheses for context. Include at least the last 3–5 verbatim exchanges.
call_type	enum	No	The type of call context (e.g., 'base'). Defaults to 'base'.
current_user	string	No	The name of the person requesting the analysis. Recommendations will be addressed to this person.

Example Invocation

{
  "name": "analyze_transcript",
  "arguments": {
    "transcript": "Customer: I just don't think Friday is realistic.\nAgent: What if we cut the scope instead?\nCustomer: That could work actually.",
    "current_user": "Agent"
  }
}

Output

Returns a JSON object detailing the detected emotional shifts, underlying dynamics, and actionable recommendations for the current_user to steer the conversation effectively.

---

get_credits

Checks the remaining analysis credits for the authenticated user for the current billing period.

Inputs

This tool takes no inputs.

Example Invocation

{
  "name": "get_credits",
  "arguments": {}
}

Output

Returns a JSON string containing your available credits and their expiration date:

{
  "available_credits": 950,
  "credits_expire_at": "2026-08-01T00:00:00.000Z"
}