MCP server for the FIVE Persona Engine — an LLM persona constraint engine that generates structured JSON constraints to eliminate persona drift. Instead of describing personality in words (which LLMs interpret differently each turn), FIVE defines behavioral parameters the LLM executes as a recipe. See how it works →
Measured: with the constraint JSON + free harness, the demo character survived a 120-turn pressure test with zero persona breaks (plain prompt: 8 breaks; JSON alone: 1). Numbers, transcripts and scripts →
Free — no API key, no account.
pip install five-mcpAdd to your claude_desktop_config.json:
{
"mcpServers": {
"five-character-engine": {
"command": "five-mcp"
}
}
}That's it — no environment variables needed.
Any MCP-compatible client can connect via stdio transport:
five-mcpRelated MCP server: CSL-Core
generateGenerates persona constraints via the FIVE engine.
Parameter | Type | Required | Description |
| string | Yes | Name of the character |
| A / B / C / D | Yes | Personality axis choices |
| 1–5 | No | Style sliders (default: 3) |
| string | No | Free-form description |
{
"status": "ok",
"constraint": { "..." }
}The API is free. To keep it available to everyone, a light per-IP limit applies: 10 requests/min, 200 requests/day. The free host may cold-start, so the first call can take up to a minute.
API & Docs: fiveengine.dev
GitHub: github.com/kiro0x/five-mcp
Design philosophy & examples: five-character-engine README
Measured evaluation (120-turn drift test): five-character-engine eval/
Engine repo: github.com/kiro0x/five-character-engine
MIT
mcp-name: io.github.kiro0x/five-mcp
Generate persona constraints using the FIVE engine.
This tool calls the FIVE Persona Engine API to produce JSON constraints that prevent persona drift and keep an LLM character's voice consistent.
Each call costs $1 and consumes one credit from your account.
Args: character_name: Name of the character to generate constraints for. q1: Personality axis 1 – choose A, B, C, or D. q2: Personality axis 2 – choose A, B, C, or D. q3: Personality axis 3 – choose A, B, C, or D. q4: Personality axis 4 – choose A, B, C, or D. s1: Style slider 1 (1-5, default 3). Optional fine-tuning. s2: Style slider 2 (1-5, default 3). Optional fine-tuning. s3: Style slider 3 (1-5, default 3). Optional fine-tuning. s4: Style slider 4 (1-5, default 3). Optional fine-tuning. free_text: Optional free-form description to further guide generation.
Returns: A dict with keys: status, remaining (credits left), constraint (the generated JSON constraint object).
Inputs: character_name (required), q1 (required), q2 (required), q3 (required), +6 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 1 tool
With only one tool, there is no possibility of confusion between tools. The single 'generate' tool has a clearly distinct purpose.
The tool name 'generate' is a single verb, which is clear and follows a common convention. While there is no noun to form a verb_noun pattern, the name is consistent as the only tool.
The server has only one tool, which feels thin for a general-purpose utility. However, for a very focused single-API function, it is borderline acceptable.
The tool covers the core functionality of generating persona constraints with many parameters. Minor gaps exist (e.g., no credit management or constraint listing), but the tool is complete for its stated purpose.
A profile you control: gives your AI your role, projects and preferences, across ChatGPT and Claude.
Jailbreak-proof AI guardrails. Automated Reasoning SMT solver, not an LLM. ZK proofs included.
Deterministic JSON repair, validate, example-gen, schema-coerce for agents. Zero LLM, sub-10ms.
Memory, reflection and safety for AI agents. Hosted or local, no vector database.
