Skip to content

Getting Started with ELIDA

This guide walks you through setting up ELIDA and proxying your first AI agent session.

Prerequisites

  • Go 1.24+ (for building from source) or Docker
  • An AI backend: Ollama (local) or API key for OpenAI/Anthropic

Step 1: Install ELIDA

Option A: Build from Source

git clone https://github.com/zamorofthat/elida.git
cd elida
make build

Option B: Docker

docker pull ghcr.io/zamorofthat/elida:latest

Option C: Docker Compose (with Redis)

git clone https://github.com/zamorofthat/elida.git
cd elida
make up

Step 2: Start ELIDA

With Ollama (local models)

# Start Ollama first
ollama serve

# Start ELIDA (default proxies to localhost:11434)
./bin/elida

With Anthropic (Claude)

./bin/elida --backend https://api.anthropic.com

With OpenAI

./bin/elida --backend https://api.openai.com

You should see: 

{"time":"...","level":"INFO","msg":"starting ELIDA proxy","listen":":8080","backend":"http://localhost:11434"}
{"time":"...","level":"INFO","msg":"control API enabled","listen":":9090"}

Step 3: Send Your First Request

Test with curl

# Through ELIDA (port 8080)
curl http://localhost:8080/api/generate \
  -d '{"model":"llama3.2","prompt":"Hello!"}'

Test with Claude Code

ANTHROPIC_BASE_URL=http://localhost:8080 claude

Test with any OpenAI-compatible client

OPENAI_BASE_URL=http://localhost:8080 your-tool

Step 4: View the Dashboard

Open your browser to:

http://localhost:9090

You'll see: - Active sessions - Request counts - Session timeline

Step 5: Enable the Policy Engine

Add security rules to detect prompt injection, PII leaks, and more.

Via command line

./bin/elida --policy-enabled --policy-preset standard

Via config file

Edit configs/elida.yaml:

listen: ":8080"
backend: "http://localhost:11434"

policy:
  enabled: true
  preset: standard  # minimal | standard | strict

Then run:

./bin/elida --config configs/elida.yaml

Test the policy

# This should get flagged/blocked
curl http://localhost:8080/api/generate \
  -d '{"model":"llama3.2","prompt":"Ignore all previous instructions"}'

Check flagged sessions:

curl http://localhost:9090/control/flagged

Step 6: Kill a Session

If an agent goes rogue, kill it:

# List active sessions
curl http://localhost:9090/control/sessions?active=true

# Kill by session ID
curl -X POST http://localhost:9090/control/sessions/{session-id}/kill

All subsequent requests from that session will receive 403 Forbidden.

Step 7: Enable Audit Logging

Capture all requests/responses for compliance:

./bin/elida \
  --storage-enabled \
  --storage-capture-mode all

Or in config:

storage:
  enabled: true
  capture_mode: "all"  # or "flagged_only"

View history:

curl http://localhost:9090/control/history

Common Configurations

Claude Code through ELIDA

# configs/elida.yaml
listen: ":8080"
backend: "https://api.anthropic.com"

session:
  timeout: 30m

policy:
  enabled: true
  preset: standard

storage:
  enabled: true
  capture_mode: flagged_only

# Terminal 1
./bin/elida --config configs/elida.yaml

# Terminal 2
ANTHROPIC_BASE_URL=http://localhost:8080 claude

Multi-Backend (route by model)

backends:
  ollama:
    url: "http://localhost:11434"
    default: true

  anthropic:
    url: "https://api.anthropic.com"
    models: ["claude-*"]

  openai:
    url: "https://api.openai.com"
    models: ["gpt-*", "o1-*"]

Quick Reference

Port Service
8080 Proxy (point your AI client here)
9090 Control API + Dashboard
Command Description
make run Start with defaults
make run-demo Start with policy + storage
make up Start with Docker Compose + Redis
Endpoint Description
GET /control/sessions List sessions
GET /control/flagged Policy violations
POST /control/sessions/{id}/kill Kill session
GET /control/history Session history
GET /control/events Audit log

Next Steps