FeaturesArchitectureFor teamsOpen sourceView on GitHub
Open source and ready to self-host

The AI agent harness
you actually control.

A thoughtful interface for building, running, and sharing AI agents—on your infrastructure, with your models, your way.

Start building View source
No provider lock-in · OpenAI & Anthropic compatible · Apache-2.0
Codeline · Acme Engineering
Auth refactor × API tests ×＋
Refactor authentication middlewareCode agent · claude-3-7-sonnet
Running
ME
You 10:42

Refactor the authentication middleware to support OIDC providers, while preserving the current session API.

C
Codeline 10:42

I’ll inspect the existing middleware and session boundaries before implementing the provider abstraction.

⌕ Explore repository ✓ Completesrc/auth/middleware.ts · src/auth/session.ts · 8 files
C
Codeline 10:43

The session API is already isolated. I’m adding a protocol-neutral identity provider interface and an OIDC adapter.

⌁ Edit files 4 changes+ providers/oidc.ts + providers/types.ts ~ middleware.ts
Message Codeline…
＋ Code agent · Auto context ↑
Built for control

Everything you need.
Nothing you don't.

Codeline combines a polished agent experience with the transparency and flexibility of infrastructure you own.

Open by design

Review every line, modify any behavior, and contribute upstream. Your agent harness should never be a black box.

Self-host anywhere

Keep code, prompts, credentials, and conversations on infrastructure you trust—from a laptop to your private cloud.

Tabs that keep up

Run parallel conversations without losing context. Pin, reorder, inspect, and move quickly between active agents.

Made for organizations

Share agents and runs, manage members, and turn effective workflows into durable team knowledge.

Enterprise identity

Connect existing identity infrastructure through SSO and OIDC, with centralized access for every workspace.

Deep agent composition

Customize subagents, tools, instructions, context, and handoffs to match the exact shape of your workflows.

Your infrastructure, your models

Portable from the first prompt.

Codeline's client-server architecture is designed to run where the work happens: securely on your server, close to your repositories and services.

  • Any model from any OpenAI-compatible provider
  • Native compatibility with Anthropic protocol
  • Change configuration and reload without restarts
  • Zero dependency on a proprietary control plane
ClientsBeautiful web UI

Fast conversations, tabs, sharing, and administration.

Your serverCodeline runtime

Agents, tools, organization data, and provider routing.

OpenAI protocol
Anthropic protocol
Local models
Adapt without interruption

Shape agents around
how your team works.

Build focused subagents with their own models, tools, prompts, and permissions—then update them while Codeline keeps running.

Configuration that stays live

Edit agent definitions and provider routes with confidence. Codeline reloads changes automatically, so experiments stay quick and production stays online.

Hot reloadCustom toolsModel routingSubagents
# agents/reviewer.yaml
name: code-reviewer
model: anthropic/claude-sonnet
instructions: |
Review changes for correctness, security,
maintainability, and test coverage.

subagents:
- security-auditor
- test-planner

tools: [read, search, diff]

✓ Configuration changed
✓ Reloaded code-reviewer in 38ms
No restart required

Build with agents.
Keep the control.

Deploy Codeline on your infrastructure and give your team an AI workspace built to last.

Get Codeline Read the docs