maverick-mcp
Every Goose needs its Maverick.
A behavioral intelligence MCP server that makes AI agents adapt to who they're working with — not just what they're asked to do.
Built for Goose. Works with any MCP-compatible agent.
What it does
Maverick observes how you interact with your AI agent and adapts the agent's behavior to match your style:
- Fast movers get short, decisive responses with maximum autonomy
- Methodical thinkers get thorough explanations with verification checkpoints
- Team players get collaborative framing with alignment checks
- Detail-oriented builders get precise, evidence-backed responses with process gates
Same model underneath. Different teammate on top.
How it works
┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ You │────▶│ Maverick │────▶│ Goose │
│ (developer) │ │ (MCP server)│ │ (AI agent) │
└──────────────┘ └──────────────┘ └──────────────┘
│ Detects your │
│ behavioral │
│ profile from │
│ interaction │
│ patterns │
│ │
│ Injects │
│ adaptation │
│ context into │
│ agent prompts │
└───────────────┘
- First use: Quick 2-minute behavioral assessment (10 questions) or skip and let Maverick infer from your interactions
- Profile detection: Maps you to one of 17 behavioral profiles based on four drives: Dominance, Extraversion, Patience, Formality
- Adaptation: Injects behavioral context into the agent's system prompt — adjusting tone, pacing, autonomy level, and decision style
- Learning: Refines your profile over time based on what you accept, reject, and how you interact
Quick start
# Install
pip install maverick-mcp
# Run the MCP server
maverick serve
# Or with uvx (no install needed)
uvx maverick-mcp serve
Connect to Goose
Add to your Goose configuration:
extensions:
maverick:
type: stdio
cmd: maverick
args: [serve]
Or if running as a standalone server:
extensions:
maverick:
type: sse
uri: http://localhost:8300/sse
MCP Tools
Maverick exposes these tools to the agent:
| Tool | Description |
|---|---|
maverick_assess |
Run the behavioral assessment (10 questions) |
maverick_profile |
Get the current user's behavioral profile |
maverick_adapt |
Get adaptation instructions for the current context |
maverick_feedback |
Record user feedback to refine the profile |
The 17 Profiles
Based on Predictive Index behavioral science, used by 10,000+ companies:
| Profile | Style | Agent Adaptation |
|---|---|---|
| Maverick | Fast, bold, breaks molds | Short responses, max autonomy, ship dates |
| Captain | Commanding, decisive | Executive summaries, clear delegation |
| Persuader | Energetic, narrative-driven | Compelling framing, stakeholder angles |
| Venturer | Independent, gut-driven | Minimal hand-holding, trust their direction |
| Strategist | Analytical, big-picture | Systems thinking, long-range implications |
| Controller | Structured, demanding | Precise specs, clear ownership |
| Promoter | Social, fast-moving | High energy, visible progress |
| Adapter | Balanced, flexible | Moderate everything, read the room |
| Collaborator | Team-oriented, supportive | Alignment checks, inclusive language |
| Individualist | Self-reliant, focused | Respect their process, don't over-explain |
| Artisan | Craftsmanship, patient | Quality focus, tangible outcomes |
| Guardian | Thorough, protective | Evidence first, risk assessment, checkpoints |
| Specialist | Deep, precise | Domain depth, highest standards |
| Analyzer | Methodical, data-driven | Show the data, verify claims |
| Scholar | Knowledge-seeking, rigorous | Research backing, source verification |
| Operator | Reliable, steady | Clear steps, predictable cadence |
| Altruist | Warm, caring | Supportive tone, team-aware framing |
Data sovereignty
Your behavioral profile is stored locally on your machine — never sent to any server. Maverick runs entirely local. No accounts, no telemetry, no cloud.
~/.maverick/
├── profile.yaml # Your behavioral profile
├── interactions.jsonl # Interaction history (for refinement)
└── config.yaml # Server configuration
Built on
- Predictive Index behavioral science framework
- Model Context Protocol (MCP) for agent integration
- Airlock behavioral intelligence platform
The science
The four behavioral drives (DECF):
- Dominance — Drive for control and influence over environment
- Extraversion — Drive for social interaction and influence over people
- Patience — Drive for consistency, stability, and deliberate pace
- Formality — Drive for conformity, rules, and structure
Every person has a unique DECF signature. Every AI interaction should respect it.
We've validated this across 7,000+ conversations using ConstellationBench — a novel persona fidelity benchmark that measures whether AI agents maintain consistent behavioral adaptation.
Why "Maverick"?
The Maverick is the rarest behavioral profile — 2.5% of the population. Independent innovators who break every mold and bet on the future others can't see.
We named this project after them because that's what's missing from AI agents: the ability to recognize that not everyone thinks the same way, and to adapt accordingly.
Every Goose needs its Maverick.
License
Apache 2.0 — same as Goose.
Contributing
PRs welcome. See CONTRIBUTING.md.
Built by Airlock — the AI that knows who it's talking to.
Metadata
Release files for maverick-mcp 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| maverick_mcp-0.1.0.tar.gz | 75.0 kB | Without provenance | Details for maverick_mcp-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| maverick_mcp-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for maverick_mcp-0.1.0-py3-none-any.whl |
Total release size: 97.7 kB
Release files / maverick_mcp-0.1.0.tar.gz
| Download URL | maverick_mcp-0.1.0.tar.gz |
|---|---|
| Size | 75.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
ac34f375526e4913876400ce5137fc2b46f930810586b55f963cd14591f0fc0a
|
|
BLAKE2b-256 checksum How to use checksums |
8acfb0a7f8dfa8d0bb76f6d096ad8e95be0dcc8169ee011131be6172a645a416
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.10.10 {"installer":{"name":"uv","version":"0.10.10","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / maverick_mcp-0.1.0-py3-none-any.whl
| Download URL | maverick_mcp-0.1.0-py3-none-any.whl |
|---|---|
| Size | 22.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
8438708a8da4f146128768f8bf6dccf0d9d8cfb7de761032be9145ccdabd75b1
|
|
BLAKE2b-256 checksum How to use checksums |
79623bcc41a56512f0fe5d1ca23258d9c283f997d80988d82917c96ac9020180
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.10.10 {"installer":{"name":"uv","version":"0.10.10","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Security for maverick-mcp 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.