Laxmi Agents
Open-source agentic framework for revenue intelligence
6 AI agents that extract, resolve, compute, detect, and explain your business data.
What is Laxmi Agents?
Laxmi Agents (PyPI: vesh-agents) is an agentic AI framework for business data analysis, starting with SaaS revenue intelligence. Built on the OpenAI Agents SDK, it provides:
- 6 specialized agents that work together as a pipeline
- BYOM (Bring Your Own Model) — use Claude, DeepSeek, GPT, Llama, or any LiteLLM-compatible model
- Entity resolution — automatically match records across Stripe, Postgres, CSV, and more
- SaaS metric computation — MRR, churn, ARPU, NRR, Quick Ratio out of the box
- Anomaly detection — statistical methods find unusual patterns in your metrics
- CLI + MCP server — use from terminal or integrate with Cursor, OpenCode, Claude Desktop
Quickstart
pip install vesh-agents
Analyze a CSV in 30 seconds
Use with any LLM (BYOM)
from agents import Runner
from vesh_agents.verticals.revenue import create_revenue_orchestrator
# Use Claude, DeepSeek, GPT, or any LiteLLM model
orchestrator = create_revenue_orchestrator(model="litellm/anthropic/claude-sonnet-4-20250514")
result = await Runner.run(
orchestrator,
"Analyze revenue from examples/sample_data.csv. Why is churn increasing?"
)
print(result.final_output)
Connect to live Stripe
export STRIPE_API_KEY=sk_live_...
vesh analyze stripe
Architecture
┌──────────────────────────────────────────────────────────────┐
│ Vesh Orchestrator Agent │
│ Coordinates the pipeline based on natural language queries │
└──────────┬───────────┬──────────┬──────────┬────────────────┘
│ │ │ │
┌──────▼──┐ ┌─────▼───┐ ┌───▼────┐ ┌──▼──────┐ ┌────────┐
│ Data │ │ Entity │ │ Metric │ │Anomaly │ │Insight │
│Connector│→ │Resolver │→│Computer│→│Detector│→│Reasoner│
└─────────┘ └─────────┘ └────────┘ └────────┘ └────────┘
│ │ │ │ │
CSV/Stripe/ Blocking MRR/Churn Z-score BYOM LLM
Postgres Scoring ARPU/NRR Rate-of- Explanation
Clustering Quick Ratio change
Each agent has its own tools and instructions. The orchestrator delegates to specialists via handoffs. All agents use your chosen LLM model.
Available Agents
| Agent | Role | MCP Tools |
|---|---|---|
| DataConnector | Extract data from sources | import_csv, extract_stripe, extract_postgres |
| EntityResolver | Match records across sources | resolve_entities |
| MetricComputer | Compute SaaS metrics | compute_metrics, list_metrics |
| AnomalyDetector | Find statistical anomalies | detect_anomalies |
| InsightReasoner | Explain root causes | explain_anomaly |
| Vesh Orchestrator | Coordinate the pipeline | analyze_csv (full pipeline) |
SaaS Metrics Computed
| Metric | Description |
|---|---|
| MRR | Monthly Recurring Revenue |
| New MRR | Revenue from new subscriptions |
| Expansion MRR | Revenue increase from upgrades |
| Contraction MRR | Revenue decrease from downgrades |
| Churned MRR | Revenue lost from cancellations |
| NRR | Net Revenue Retention |
| Active Customers | Count of active subscriptions |
| ARPU | Average Revenue Per User |
| Quick Ratio | Growth efficiency ratio |
| Logo Churn Rate | Customer loss percentage |
CLI Reference
# Quick offline analysis (no LLM needed)
vesh analyze csv revenue.csv
vesh analyze stripe # reads STRIPE_API_KEY from env
vesh analyze postgres --host db.example.com --database myapp # reads PGUSER/PGPASSWORD from env
# Interactive AI chat (powered by OpenCode)
vesh setup # one-time: install OpenCode + configure MCP
vesh chat # open the TUI analyst
vesh chat --model anthropic/claude-sonnet-4-20250514
# Natural language analysis (requires LLM)
vesh run "Why did churn spike last week?" --source csv:revenue.csv
# Output formats
vesh analyze csv data.csv --output json # JSON to stdout
vesh analyze csv data.csv --output rich # Rich terminal (default)
# MCP server (for Cursor, Claude Desktop, or other MCP clients)
vesh mcp serve
Using as a Python Library
# Direct pipeline (no LLM needed)
from vesh_agents.connectors.csv import CsvConnector
from vesh_agents.metrics.computation import MetricComputationEngine
from datetime import date
connector = CsvConnector(connection_id="demo", config={"file_path": "data.csv"})
records = await connector.extract_full()
entities = [r.data for r in records]
engine = MetricComputationEngine()
metrics = engine.compute_all("tenant", date.today(), entities)
for m in metrics:
print(f"{m.metric_id}: {m.value}")
Verticals
Vesh Agents uses a vertical architecture. Each vertical packages domain-specific agents, prompts, and metrics for a particular business use case.
Revenue Intelligence (included) — SaaS metrics, churn analysis, revenue decomposition
More verticals coming soon. Or build your own:
from vesh_agents.core.vertical import Vertical, VerticalConfig
class CustomerSuccess(Vertical):
config = VerticalConfig(
name="customer_success",
description="Customer health scoring and churn prediction",
metric_ids=["health_score", "nps", "usage_frequency"],
)
MCP Integration
Vesh Agents ships a real MCP server (FastMCP, stdio transport) exposing 6 tools: analyze_csv, import_csv, compute_metrics, resolve_entities, detect_anomalies, list_metrics.
OpenCode (recommended)
vesh setup # writes opencode.json + .opencode/agents/
vesh chat # launches OpenCode TUI with the Vesh analyst agent
Cursor / Claude Desktop
Add to your MCP config (.cursor/mcp.json or claude_desktop_config.json):
{
"mcpServers": {
"vesh": {
"command": "vesh",
"args": ["mcp", "serve"]
}
}
}
Manual
vesh mcp serve # starts stdio MCP server — any MCP client can connect
Vesh AI Cloud (Optional)
The open-source framework runs entirely locally. For teams that want managed infrastructure:
- Automated daily pipelines — connect once, get daily insights
- Cross-company benchmarks — "Your NRR is 40th percentile for Series B SaaS"
- Historical intelligence — 90-day trends and institutional memory
- Slack/Teams delivery — daily revenue briefs, anomaly alerts
- Agent Console — visual dashboard for agent execution traces
Visit vesh-ai.netlify.app to learn more.
Contributing
We welcome contributions! See CONTRIBUTING.md for setup instructions, coding standards, and areas where help is needed.
git clone https://github.com/shaileshai/Vesh-AI.git
cd Vesh-AI/vesh-agents
pip install -e ".[dev]"
pytest # 83 tests
License
Apache 2.0 — see LICENSE for details.
Built with ❤️ by Vesh AI
Metadata
Release files for vesh-agents 0.1.3
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 |
|---|---|---|---|---|
| vesh_agents-0.1.3.tar.gz | 477.8 kB | Without provenance | Details for vesh_agents-0.1.3.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| vesh_agents-0.1.3-py3-none-any.whl | Python 3 | none | any | Details for vesh_agents-0.1.3-py3-none-any.whl |
Total release size: 540.2 kB
Release files / vesh_agents-0.1.3.tar.gz
| Download URL | vesh_agents-0.1.3.tar.gz |
|---|---|
| Size | 477.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
dceb3677551bf8bdfd6f0f4ca92f439634165fe23296ee454aed80f9630a370d
|
|
BLAKE2b-256 checksum How to use checksums |
e0da18ef9be11cadf83ee773b4db31fb4458d8236e00fb22129c4a615cd8084d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.15
|
Release files / vesh_agents-0.1.3-py3-none-any.whl
| Download URL | vesh_agents-0.1.3-py3-none-any.whl |
|---|---|
| Size | 62.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
727c6966d38905dc826d657319a0e1902ae80a084d47701c73fe2ec0dd18a376
|
|
BLAKE2b-256 checksum How to use checksums |
6c0cf331770480dc8143b8c34309f1aa23baae74fcf2b0f8fda16cfc0ae7933c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.15
|
Security for vesh-agents 0.1.3
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.