Integrations
Connect any agent stack to Solen: LangGraph, CrewAI, AutoGen, custom agents, and your existing repos.
Solen instruments agents wherever they run. Use the SDK for in-process telemetry, webhooks for custom signals, or the GitHub App to monitor agent codebases and open heal PRs.
- GitHub App →
Monitor agent repos, PR risk analysis, and heal PRs on failure.
- @solenai/sdk →
Quick start and full config for
@solenai/sdk.
| Channel | Best for | Install |
|---|---|---|
| @solenai/sdk | LangGraph, CrewAI, AutoGen, custom Node agents | npm install @solenai/sdk |
| Python SDK | Python agent frameworks | pip install solen-sdk |
| Webhook | Any language, any agent runtime | Configure in Settings |
| GitHub App | Agent repos with production deployments | GitHub Marketplace |
All channels feed into the same Runtime Kernel heal pipeline. A confirmed agent failure triggers diagnosis and, if autonomy mode allows, automatic checkpoint resume.
@solenai/sdk
Full reference: SDK guide.
Install:
npm install @solenai/sdkSetup (3 lines):
import { withSolen } from '@solenai/sdk';
await withSolen(
{ apiKey: process.env.SOLEN_API_KEY!, agentId: 'my-agent', framework: 'langgraph' },
async (ctx) => { /* your agent logic */ }
);Get your API key from Settings → API Keys.
What it captures: run lifecycle, every step (tool calls, LLM calls), checkpoints, token usage, and step failures.
Manual checkpoint:
import { withSolen } from '@solenai/sdk';
await withSolen(config, async (ctx) => {
const state = await runPlanner();
await ctx.checkpoint(state);
return state;
});The SDK never throws or crashes your agent. Silent on network failure.
Python SDK
Install:
pip install solen-sdkSetup:
from solen_sdk import with_solen
@with_solen(api_key="...", agent_id="my-agent", framework="crewai")
async def run_agent(ctx):
result = await ctx.step("tool_call", search_web, tool_name="search_web")
await ctx.checkpoint({"result": result})
return resultWhat it captures: same telemetry as the Node SDK — runs, steps, checkpoints, and failures.
Webhooks
Send agent failure signals from any runtime in any language.
Create an endpoint: Settings → Integrations → Webhooks → New Endpoint. Solen generates a unique URL and signing secret.
Send a signal:
curl -X POST https://api.solenai.ca/api/webhooks/incident/<endpointId> \
-H "Content-Type: application/json" \
-H "X-Solen-Signature: <hmac-sha256-of-body>" \
-d '{
"agentId": "research-agent",
"runId": "clx9k2m00000008l4abc123de",
"error": "tool_call_loop: search_web x23",
"stepNumber": 47,
"timestamp": "2026-05-17T03:42:08Z"
}'Signing requests: compute HMAC-SHA256 of the raw request body using your secret:
import { createHmac } from 'crypto'
const sig = createHmac('sha256', secret).update(rawBody).digest('hex')import hmac, hashlib
sig = hmac.new(secret.encode(), raw_body.encode(), hashlib.sha256).hexdigest()Outbound heal events: configure a URL to receive heal lifecycle notifications:
| Event | Fired when |
|---|---|
heal.triggered | Agent failure rule fires (stall, tool loop, step error) |
heal.proposed | Fix ready, awaiting approval (supervised mode) |
heal.applying | Heal approved, resuming from checkpoint |
heal.resolved | Agent step activity restored and verified |
heal.failed | Heal did not restore step activity after resume |
Vercel
Connect Vercel from Settings → Integrations → Vercel. Solen correlates agent deployment events with heal incidents in your release pipeline.
Slack
Install the Solen Slack app from Settings → Integrations → Slack to receive heal approval requests and agent incident alerts in your team channel.
Docker Sidecar
Run the solenai/agent sidecar alongside your agent container. It forwards step heartbeats and failure signals to the Runtime Kernel when in-process SDK instrumentation is not possible.
docker run -d --name solen-agent \
-e SOLEN_AGENT_TOKEN=YOUR_AGENT_TOKEN \
-e SOLEN_TARGET_CONTAINER=my-agent \
-e SOLEN_AGENT_ID=research-agent \
-v /var/run/docker.sock:/var/run/docker.sock:ro \
--network container:my-agent \
gautamtalksdev/agent:v3Works on Docker hosts and Kubernetes pods (mount the Docker socket or use a shared network namespace).
GitHub App
Monitor any agent repository with a production deployment, not just agents instrumented via SDK. See the GitHub App guide for installation, PR analysis, and linking repos to agents.
Install: Find "Solen" on the GitHub Marketplace and click Install. Select agent repositories to monitor.
Configure: After install, go to Settings → Integrations → GitHub App. For each connected repo, set:
- Linked
agentIdfor the agent deployed from this repo - Autonomy mode (
supervisedorautonomous) - Optionally a branch for heal PRs
What Solen does:
- Monitors agent run health via SDK telemetry or webhooks
- On agent failure, creates an incident and diagnoses root cause
- In supervised mode: opens a PR with the proposed fix
- In autonomous mode: applies the heal and opens a PR showing what changed