@arkheia/mcp-server 0.1.2 → 0.1.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +43 -44
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -1,21 +1,18 @@
1
- # Arkheia MCP Server — Fabrication Detection for LLMs
1
+ # Arkheia MCP Server — Fabrication Detection for AI Agents
2
2
 
3
- Detect fabrication (hallucination) in any LLM output.
4
- Free tier included (1,500 detections/month).
3
+ Know when your AI is making things up.
5
4
 
6
- ## Quick Start
5
+ Arkheia screens model responses for fabrication using behavioural fingerprinting. Works with Claude, GPT, Gemini, Grok, Llama, Mistral, and 30+ other models. One tool call. Real-time risk scoring.
6
+
7
+ Free tier: 1,500 detections/month. No credit card.
7
8
 
8
- ### 1. Clone and install
9
+ ## Quick Start
9
10
 
10
11
  ```bash
11
- git clone https://github.com/arkheiaai/arkheia-mcp.git ~/.arkheia-mcp
12
- cd ~/.arkheia-mcp
13
- python -m venv .venv
14
- source .venv/bin/activate # Windows: .venv\Scripts\activate
15
- pip install -r requirements.txt
12
+ npx @arkheia/mcp-server
16
13
  ```
17
14
 
18
- ### 2. Get an API key (free)
15
+ Get a free API key:
19
16
 
20
17
  ```bash
21
18
  curl -X POST https://arkheia-proxy-production.up.railway.app/v1/provision \
@@ -23,68 +20,70 @@ curl -X POST https://arkheia-proxy-production.up.railway.app/v1/provision \
23
20
  -d '{"email": "you@example.com"}'
24
21
  ```
25
22
 
26
- ### 3. Add to Claude Desktop
27
-
28
- Edit your config file:
29
- - macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
30
- - Windows: `%APPDATA%\Claude\claude_desktop_config.json`
23
+ Add to your agent config (Claude Code, Claude Desktop, Cursor, or any MCP-compatible tool):
31
24
 
32
25
  ```json
33
26
  {
34
27
  "mcpServers": {
35
28
  "arkheia": {
36
- "command": "~/.arkheia-mcp/.venv/bin/python",
29
+ "command": "python",
37
30
  "args": ["-m", "mcp_server.server"],
38
- "cwd": "~/.arkheia-mcp",
31
+ "cwd": "~/.arkheia/mcp",
39
32
  "env": {
40
- "PYTHONPATH": "~/.arkheia-mcp",
41
- "ARKHEIA_API_KEY": "ak_live_YOUR_KEY_HERE"
33
+ "PYTHONPATH": "~/.arkheia/mcp",
34
+ "ARKHEIA_API_KEY": "ak_live_your_key_here"
42
35
  }
43
36
  }
44
37
  }
45
38
  }
46
39
  ```
47
40
 
48
- On Windows, replace `~/.arkheia-mcp` with the full path using forward slashes (e.g. `C:/Users/YourName/.arkheia-mcp`) and use `.venv/Scripts/python` instead of `.venv/bin/python`.
49
-
50
- ### 4. Restart Claude and verify
51
-
52
- Restart Claude Desktop (quit and reopen, not just close window). The Arkheia tools will appear automatically.
41
+ Restart your agent. Then ask it:
53
42
 
54
- To verify, ask Claude: *"Use arkheia_verify to check this response for fabrication: 'The Eiffel Tower is located in Berlin, Germany.'"*
43
+ > "Use arkheia_verify to check this response: The Kafka 4.1 ConsumerLease API introduces a lease-based partition ownership model."
55
44
 
56
- You should see a detection result with a risk level and confidence score.
45
+ It should flag this as **HIGH** risk — because the Kafka 4.1 ConsumerLease API doesn't exist.
57
46
 
58
- ## Tools
47
+ ## What You Get
59
48
 
60
49
  | Tool | Description |
61
50
  |------|-------------|
62
- | `arkheia_verify` | Score any model output for fabrication risk (LOW/MEDIUM/HIGH) |
63
- | `arkheia_audit_log` | Review detection history |
64
- | `run_grok` | Call xAI Grok + screen for fabrication |
65
- | `run_gemini` | Call Google Gemini + screen for fabrication |
66
- | `run_together` | Call Together AI (Kimi, DeepSeek) + screen |
51
+ | `arkheia_verify` | Score any model response for fabrication risk (LOW/MEDIUM/HIGH) |
52
+ | `arkheia_audit_log` | Review your detection history |
53
+ | `run_grok` | Call Grok + screen for fabrication |
54
+ | `run_gemini` | Call Gemini + screen for fabrication |
67
55
  | `run_ollama` | Call local Ollama model + screen |
68
- | `memory_store` / `memory_retrieve` / `memory_relate` | Persistent knowledge graph |
56
+ | `run_together` | Call Together AI (Kimi, DeepSeek) + screen |
57
+
58
+ ## 35+ Model Profiles
59
+
60
+ GPT-4o, GPT-5.4, Claude Opus/Sonnet/Haiku, Gemini 2.5/3.0, Grok 4, Llama, Mixtral, CodeLlama, Falcon, Phi4, Kimi K2.5, and more. If your model isn't listed, [let us know](mailto:dmurfet@arkheia.ai).
69
61
 
70
62
  ## Pricing
71
63
 
72
- - **Free:** 1,500 detections/month (no credit card)
73
- - **Single Contributor:** $99/month (unlimited)
74
- - **Professional:** $499/month (20 concurrent)
75
- - **Team:** $1,999/month (50 concurrent)
64
+ | Plan | Price | Detections |
65
+ |------|-------|------------|
66
+ | Free | $0 | 1,500/month |
67
+ | Single Contributor | $99/month | Unlimited |
68
+ | Professional | $499/month | Unlimited |
69
+ | Team | $1,999/month | Unlimited |
70
+
71
+ ## Full Documentation
76
72
 
77
- ## Requirements
73
+ See the [GitHub repo](https://github.com/arkheiaai/arkheia-mcp) for:
74
+ - Complete setup guide for all agents
75
+ - CLAUDE.md template for automatic detection across projects
76
+ - Multi-agent quorum pattern
77
+ - Test prompts and examples
78
78
 
79
- - Python 3.10+
80
- - Git
79
+ ## Feedback
81
80
 
82
- ## Full Setup Guide
81
+ - **GitHub Issues:** https://github.com/arkheiaai/arkheia-mcp/issues
82
+ - **Email:** dmurfet@arkheia.ai
83
83
 
84
- See [AGENTS.md](https://github.com/arkheiaai/arkheia-mcp/blob/master/AGENTS.md) for detailed instructions, troubleshooting, and environment variables.
84
+ Every message is read by the founder.
85
85
 
86
86
  ## Links
87
87
 
88
88
  - Website: https://arkheia.ai
89
89
  - GitHub: https://github.com/arkheiaai/arkheia-mcp
90
- - Support: support@arkheia.ai
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@arkheia/mcp-server",
3
- "version": "0.1.2",
3
+ "version": "0.1.3",
4
4
  "mcpName": "io.github.arkheiaai/mcp-server",
5
5
  "description": "Arkheia MCP Server — Fabrication detection for LLM outputs. Detect hallucination in any model's output with a single tool call.",
6
6
  "bin": {