Orchestrator Integration

Connect your agent orchestrator to the Rosentic Merge Index.

Three Tiers of Intelligence

Tier 1

Rules Feed

Fetch once at session start and inject merge-safe rules into agent prompts.

Available Now

Tier 2

Hotspots

Check current repo surfaces before assigning work to an agent.

Coming Soon

Tier 3

Scan Gate

Run check-siblings before fan-in to validate concrete branches.

CLI Available Now

Integration Guides

Conductor

Rules Feed

Conductor is a macOS app for managing multiple AI coding agents.

Fetch rules at session start, then append high-severity rules to each agent prompt.

Fetch rulescurl
curl https://api.rosentic.com/v1/feed/rules \
  -H "Authorization: Bearer $ROSENTIC_API_KEY"

Prompt injectiontext
Before editing contract-bearing code:
- If you change a function signature, update all callers in this branch.
- If you remove response fields, verify active branch consumers first.
- If another agent is touching the same entry point, wait or scope around it.

Without Rosentic

Agents can create branch-compatible changes that break when landed together.

With Rosentic

Agents receive rules targeted at stale callers, the pattern behind 87% of observed conflicts.

Claude Squad

MCP Native

Claude Squad is an open-source tool for running multiple Claude Code instances.

Use the Rosentic MCP server for investigation and check-siblings for pre-merge validation.

MCP configjson
{
  "mcpServers": {
    "rosentic": {
      "command": "rosentic-mcp",
      "args": []
    }
  }
}

Pre-merge validationbash
python3 detect.py check-siblings /path/to/repo \
  --base main \
  --branches agent/a,agent/b,agent/c \
  --format json

LangGraph

Python Wrapper

LangGraph is a framework for building stateful agent systems.

Wrap the rules feed as a tool, then run check-siblings as a pre-merge node.

Tool wrapperpython
import os
import subprocess
import requests
from langchain_core.tools import tool

@tool
def get_rosentic_rules() -> dict:
    resp = requests.get(
        "https://api.rosentic.com/v1/feed/rules",
        headers={"Authorization": f"Bearer {os.environ['ROSENTIC_API_KEY']}"},
        timeout=10,
    )
    resp.raise_for_status()
    return resp.json()

@tool
def check_conflicts(repo_path: str, branches: list[str]) -> str:
    cmd = [
        "python3", "detect.py", "check-siblings", repo_path,
        "--base", "main",
        "--branches", ",".join(branches),
        "--format", "json",
    ]
    return subprocess.check_output(cmd, text=True)

StateGraph nodepython
from langgraph.graph import StateGraph

def pre_merge_gate(state):
    result = check_conflicts.invoke({
        "repo_path": state["repo_path"],
        "branches": state["agent_branches"],
    })
    state["rosentic_scan"] = result
    return state

graph = StateGraph(dict)
graph.add_node("pre_merge_gate", pre_merge_gate)

Quick Start

Step 1: Get an API key

Create a workspace in the Rosentic dashboard and copy your API key.

Step 2: Fetch rules

Requestbash
curl https://api.rosentic.com/v1/feed/rules \
  -H "Authorization: Bearer $ROSENTIC_API_KEY"

Step 3: Inject into your agents

Agent prompttext
Use these Rosentic rules during this session:
1. Update same-branch callers after signature changes.
2. Check consumers before removing response fields.
3. Avoid parallel edits on shared entry points unless scoped.

Step 4: Add check-siblings

Before fan-in merge, scan the exact branches your orchestrator is about to land.

CLIbash
python3 detect.py check-siblings /path/to/repo \
  --base main \
  --branches agent/a,agent/b,agent/c \
  --format json

API Reference

GET /v1/feed/rules

Returns fleet-learned rules for prompt injection and task routing guidance.

Responsejson
{
  "updated": "2026-06-09",
  "scans_indexed": 2432,
  "repos_indexed": 193,
  "rules": [
    {
      "id": "l1-arity-self-update",
      "layer": "L1_signature",
      "severity": "high",
      "confidence": 0.87,
      "category": "prevention"
    }
  ]
}

GET /v1/feed/hotspots

Coming Soon

Returns repo-specific active surfaces and high-pressure files for pre-assignment checks.

Response shapejson
{
  "repo": "org/repo",
  "risk_score": 44,
  "risk_tier": "elevated",
  "hot_surfaces": [
    {
      "surface": "create_order()",
      "layer": "L1_signature",
      "file": "backend/orders.py"
    }
  ]
}

POST /v1/scan

Coming Soon

Runs a hosted sibling scan and returns the standard JSON output contract plus merge plan and fix hints.

Request shapejson
{
  "repo_url": "https://github.com/org/repo",
  "base_branch": "main",
  "branches": ["agent/a", "agent/b"],
  "mode": "siblings"
}

Error responses

Status	Error	Meaning
400	invalid_request	Malformed request or missing fields.
401	unauthorized	Missing or invalid API key.
403	plan_required	Endpoint requires a higher plan.
404	repo_not_found	Repo is not in the Merge Index.
408	scan_timeout	Scan exceeded its time budget.
429	rate_limit_exceeded	Rate limit exceeded. Retry after the reset time.
500	internal_error	Unexpected server error.
503	scan_unavailable	Hosted scan runtime unavailable.

Implementation status. Tier 1 is available as a REST endpoint. Tier 2 and hosted Tier 3 are documented as future endpoints. The local check-siblings CLI is available now.