Prismona

Connector between you and AI

Plug your personality blueprint into your agent.

A Prismona blueprint encodes your personality, and this server is how AIs and other systems read it. Plugging in means two things:

Your AI agent becomes your complement
Hand your own assistant your code and it takes on a personality calibrated to suit yours — supplying the structure, calm, candor, or grounding your blueprint suggests you benefit from, instead of a one-size-fits-all voice.
Other systems learn to work with you
Anyone's agent — a colleague's assistant, a team tool, a scheduling bot — can take a code you chose to share and adapt how it communicates and collaborates with you: the right channel, the right pace, the right amount of directness. Plus pairing reads, team composition, and your working-with-me manual, on demand.

Connect any MCP-capable assistant — Claude, ChatGPT, Cursor, OpenClaw, Hermes, Ollama and more — to the live endpoint below, and it negotiates the rest.

I.

Connect

The hosted endpoint negotiates the protocol live over Streamable HTTP — nothing to install, no account, no key:

https://prismona.io/api/mcp
Claude Code
One command:
claude mcp add --transport http prismona https://prismona.io/api/mcp
Claude.ai / Desktop
Settings → Connectors → Add custom connector → paste the endpoint URL.
Cursor, agents, anything else
Standard MCP client config:
{
  "mcpServers": {
    "prismona": { "url": "https://prismona.io/api/mcp" }
  }
}
Fully offline
The same tools ship as a local stdio server (packages/mcp in the repository) for setups where blueprint codes should never leave the machine — the hosted endpoint is stateless either way.
II.

Tools

usage_guide
The agent-facing markdown guide: which tool for which job, worked examples, binding rules — agents should call this first
decode_profile
Full structured blueprint from a share code: percentiles, uncertainty ranges, archetype blend, distinctiveness
value_brief
Core value priorities (Schwartz) from a PRSM-VAL- code — top/least values, the key tension, and how to weigh trade-offs for alignment
delegation_brief
Stated operating preferences from a PRSM-DEL- code — decision rights, cadence, options-vs-recommendation, depth, ambiguity handling, verification posture; how the owner wants work run, measured by scenario, not inferred from traits
communication_brief
Stated communication register from a PRSM-COM- code — expressiveness, preciseness, directness, questioningness, emotional expression, impression management; how to read the owner and write back
compare_values
Value congruence for two PRSM-VAL- codes — alignment score, shared priorities, and friction points (the values counterpart of compare_dyad)
profile_readings
The six applied readings — relationships, career, work style, leadership, integrity, cofounder
compare_dyad
Pairing report for two codes: fit gauge, strengths, frictions with conversation prompts
agent_handshake
Two agents representing two people exchange negotiation profiles and compute a coordination protocol (commitments, conflict, trust) — the machine-to-machine counterpart of compare_dyad
proxy_brief
A bounded-proxy operating brief for an agent acting for an absent owner: act-vs-defer gate, explicit scope/stakes/expiry, and disclose-don't-impersonate
team_composition
Two or more codes: trait diversity, role coverage, gaps, single points of failure, gates
compose_team
Forward composition: project outcome + Team Topologies shape + headcount → seat-by-seat archetype plan
compose_agents
An agent bench that complements one person: roles + voice flavors staffed around their measured blueprint
team_personas
Every calibrated persona in a published agent team, from a PRSM-TEAM- code — the live pull for orchestrators
tune_agent
Report what worked with the owner so a team agent's persona learns over time (behavioral only; folds into its learned overlay)
working_with_me
The first-person collaboration one-pager
agent_persona
Complement-calibrated system prompt, tunable by voice flavor, professional role, and relationship comportment
comportment_adapter
The register (formality, deference, disclosure…) an agent should carry for a given relationship — the persona stays fixed, only the comportment adapts
interaction_guide
Third-person guidance for communicating with the person behind a code
management_style
Questionnaire default + the weekly field-notes digest reported by agents who worked with them
report_collaboration
Submit what worked / didn't after collaborating — folds into the owner's weekly digest
submit_observation
Submit a behavioral summary (communication, work style, strategies, quirks) — feeds the owner's living observed layer, never their measured scores

Try it: connect, then ask your assistant — “Here are our two Prismona codes. Compare us as cofounders and tell me which conversation we should have first.”

III.

Consent & privacy

Every tool takes a share code as input — and a code only exists because its owner generated and handed it over. Possession is the grant; a code shared for one purpose is not consent for another. The endpoint is stateless: codes are decoded in-process, nothing is stored, and no identity is attached. Outputs are probabilistic readings with stated limits — never verdicts, and never a basis for screening anyone. Methodology · Privacy Policy