Skip to content
VENKAI
Pilot

Pilot

A pilot is for measuring, not confirming.

The scenario below is not a mock-up: it runs, and it replays in one command. What it isn't is a customer deployment — Venkai is in pilot phase, with no production installation to date.

Scenario

A four-agent chain

Four agents follow each other on the same case. None reads the others' conversations: they read a state.

  1. 01Research Agentestablishes the facts and writes the constraints of the case
  2. 02Proposal Agentreads the constraints, proposes, writes its decisions and their reasons
  3. 03Review Agentreads the decisions, writes its objections and blockers
  4. 04Implementation Agentreads decisions and objections, executes, writes progress

Each agent writes to the same project and reads by query. The Review Agent doesn't need the Research Agent's transcript: it asks for decisions and constraints, and receives six objects instead of forty pages.

Interfaces

Three ways to wire an agent

Python SDK
from venkai import VenkaiClient

client = VenkaiClient(api_key="vk_live_...", base_url="https://api.venkai.fr")

client.memory.remember(
    "Client budget: 50k EUR, hard cap",
    project="onboarding", type="constraint",
    agent="research_agent", importance=0.95,
)

memories = client.memory.recall(query="budget constraints",
                                project="onboarding", limit=5)
REST
curl -X POST https://api.venkai.fr/api/memory \
  -H "Authorization: Bearer vk_live_..." \
  -H "Content-Type: application/json" \
  -d '{"project_id":"onboarding","agent_id":"agent1",
       "content":"Budget 50k","type":"constraint"}'

curl "https://api.venkai.fr/api/context/onboarding/relevant?query=budget&limit=5" \
  -H "Authorization: Bearer vk_live_..."
MCP
{
  "venkai": {
    "command": "python",
    "args": ["venkai/mcp/server.py"],
    "env": { "VENKAI_API_KEY": "vk_live_..." }
  }
}

Scope

What a pilot can measure

These measurements don't exist yet. That is precisely the point of a pilot: they have to be taken at your site, on your workflow.

  • Actual continuity

    On a real workflow, how many resumptions happen without rebuilding state by hand.

  • Retrieval quality

    Whether what an agent receives matches what it needed in order to act.

  • Integration cost

    How many lines and write points have to be added to existing agents.

  • Effect on LLM cost

    No measurement exists today. It requires running real agents on both sides.

No result is promised before it is measured. If a pilot shows continuity wasn't your bottleneck, that is a useful result and we will say so.

Pilot

Build a continuity layer for your agents.

Describe your multi-agent workflow and where context goes missing. We answer with a pilot scope, or with the reason Venkai isn't the right piece.

No newsletter. A human reply. · or write directly to contact@venkai.fr