Coding agents, editors, and terminal tools: who makes each one, what it costs, which models it can use, and whether it can talk to hardware you own. Acquisitions, renames, and shutdowns are marked, because this field changes monthly.
Rows were last checked on 2026-09-25. Capabilities change; treat this as a starting map and confirm endpoint support against each project’s current docs before you wire anything up.
Free and open source; you pay whichever model provider you point it at.
Self-improving TUI harness launched August 5, 2026. A persistent IPython REPL is the only tool, sub-agents are function calls, and the agent can edit its own prompts, skills, and memory at runtime. Works with open and closed frontier models; a background daemon keeps sessions attachable and recoverable. It runs model-generated Python with your user permissions and ships no sandbox, so give it the same trust you would give a script you have not read.
Included in every paid Claude plan (Pro $20/month, Max from $100/month, Team, Enterprise), or pay per token with Claude API, Bedrock, Google Cloud, or Microsoft Foundry credentials.
Terminal agent (also IDE, desktop, and web) for Claude models. It officially supports custom base URLs through ANTHROPIC_BASE_URL for LLM gateways that expose a supported API format, but Anthropic does not support routing it to non-Claude models.
Included with ChatGPT Plus, Pro, Business, Edu, and Enterprise sign-in, or billed per token with an OpenAI API key.
Open-source Rust CLI that defaults to OpenAI’s GPT-6 models (gpt-6-sol at check time). Supports custom model providers by base URL and local models through Ollama or LM Studio with --oss.
Hobby (free), Individual $20/month, Teams $40/user/month, and Enterprise. Your own OpenAI, Anthropic, Google, Azure, or Bedrock keys can be added in settings.
Agentic IDE. Custom API keys are supported, but every request is routed through Cursor’s servers for prompt building, so localhost OpenAI-compatible endpoints are not counted as supported here. Cursor acquired Continue in mid-2026.
Free control surface; you pay for whichever provider CLIs you already run.
Open-source GUI and remote control plane over installed agents such as Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok, with desktop and mobile apps. It does not talk to models directly; local-endpoint support depends on the backend agent you attach.
Editor is free; built-in chat works with your own provider keys or local models, or with a GitHub Copilot plan.
Open editor shell. Built-in chat and agent mode can use Copilot’s hosted models, your own provider keys, a custom OpenAI-compatible endpoint, or local models offline without a Copilot plan; extensions such as Cline add other agents.
Free, Pro $10/month, Pro+ $39/month, Max $100/month (each with a GitHub AI Credits allowance), plus Business and Enterprise through sales. BYOK model keys work in VS Code chat.
Agent, chat, CLI, and cloud coding agent across GitHub and editors. Hosted models include Claude, GPT, and Gemini; in VS Code you can also bring your own provider key, a custom OpenAI-compatible endpoint, or a local model (organization policy must allow it on Business and Enterprise).
Nous Portal, OpenRouter, or your own keys and endpoints.
Self-improving agent with persistent memory, available as CLI, desktop app, and chat-app gateway. Provider switch includes custom endpoints; tasks can run in local, Docker, SSH, Singularity, or Modal sandboxes.
Bring hosted, subscription-backed, gateway, or local models.
Local-first personal agent gateway across chat apps, created by Peter Steinberger and now stewarded by the OpenClaw Foundation. Broader than a pure coding CLI, but commonly used to drive coding work and custom model endpoints.
Free and open source with BYOK or local models; optional OpenCode Zen pay-as-you-go and OpenCode Go ($10/month) model plans, and Copilot or ChatGPT logins.
Open coding agent for the terminal, desktop, and IDEs with first-class OpenAI-compatible provider config and 75+ providers, including local models.
Free to run; use Charm’s Hyper subscription or your own provider keys and local models.
Terminal coding agent from Charm with LSP context and MCP. Source-available under the Functional Source License (FSL-1.1-MIT, converting to MIT later), so not OSI open source yet. Supports custom OpenAI-compatible and Ollama providers.
Free CLI; the README asks for your own Mistral API key.
Minimal open-source terminal coding agent from Mistral with skills, subagents, and ACP editor integration. Models and providers are configurable in config.toml; local endpoints are not confirmed here. Mistral has also renamed Le Chat to Vibe, whose Pro plan is marketed for coding.
Open source; bring your own keys or ChatGPT/Z.ai plans, with optional Letta Cloud for hosted memory and remote computers.
Memory-first agent harness whose agents rewrite their own memory, skills, and prompts over time. Runs as CLI, desktop app, or through chat apps, against Letta Cloud or a local backend; local model endpoints are not confirmed here.
Kimi membership OAuth or Moonshot platform API keys.
Moonshot’s coding agent as a desktop app, terminal CLI, and IDE plugins (ACP). Defaults to Kimi models, including Kimi K3 on higher Kimi Code plans; docs say other compatible providers can be configured.
Free for individuals with weekly limits; Google AI Pro and Ultra raise limits; Google Cloud organizations pay by consumption. SDK is open source.
Google’s agent-first IDE, desktop agent app, CLI, and SDK. Hosted models include Gemini 3.x, Claude Sonnet/Opus 4.6, and gpt-oss-120b through Google’s stack; the SDK is open, but the product is not a self-hosted OpenAI-compatible client.
Free, Pro $20/month, Max $200/month, Teams $80/month plus $40 per seat, and Enterprise. BYOK is not listed on the current plans page.
Agentic IDE formerly called Windsurf, renamed by Cognition on June 2, 2026. Its local agent, Devin Local, replaced Cascade, and the IDE can also run ACP agents such as Codex and OpenCode. Custom self-hosted OpenAI-compatible endpoints are not treated as a supported path here.
Editor is free/open; Zed AI and third-party models bill separately.
Open editor with an agent panel that can use hosted Zed AI or custom OpenAI-compatible providers. Most source is GPL-3.0; some reusable crates are Apache-2.0.
Open-source extension is free and BYOK; the paid Continue Hub was shut down after the Cursor acquisition.
Legacy: Cursor acquired Continue in June 2026 and shut down the hosted Hub. The Apache-2.0 extension, built around configurable providers including local OpenAI-compatible servers, remains available but is no longer actively developed by the original company.
Free, Pro $20/month, Max $200/month, Teams $80/month plus $40 per seat, and Enterprise, shared with Devin Desktop.
Autonomous software engineering agent that works in a cloud environment and opens pull requests. Now part of one Devin lineup with Devin Desktop (formerly Windsurf) and a CLI.
Free, Build $20/month, Max $200/month, Business $50/user/month, Enterprise. Your own API keys and custom inference endpoints need Business or Enterprise.
Agentic terminal and development environment with a built-in agent that can also host CLI agents like Claude Code and Codex. The client has been open source under AGPL-3.0 since April 28, 2026, while the Oz cloud agent platform stays proprietary.
Hobby free, Individual $20/month (cloud orb minutes), Teams, and Enterprise. Bring your own keys or ChatGPT and other subscriptions with no Amp token fees.
Frontier coding agent from an independent lab spun out of Sourcegraph, with CLI, editor, and cloud orbs that keep running after you close your laptop. Model tokens can come from your own keys or existing subscriptions; local OpenAI-compatible endpoints are not confirmed.
Pro $20/month, Plus $100/month, Max $200/month, Teams $60/month plus $40 per seat, and Business/Enterprise. BYOK and local models are also supported.
Agent-native platform whose coding surface is Droid (CLI, desktop app, and cloud). BYOK config accepts any OpenAI-compatible base URL, including local Ollama or LM Studio servers.
Extension is free; inference is BYOK. Project is archived.
Archived: VS Code agent extension that forked from Cline with custom modes and MCP. The repo was archived in May 2026; the team now builds Roomote, a separate source-available self-hosted cloud agent, and roocode.com redirects there. Kilo Code is the maintained fork.
Free and open source with BYOK, local, or Kilo Gateway/Kilo Pass inference; Teams $15/user/month; Enterprise custom.
Open coding agent for VS Code, JetBrains, and CLI, acquired by Anaconda in July 2026 with no announced product changes. Started as a Roo Code fork and is the maintained successor now that Roo Code is archived.
Open agent; you pay whichever LLM provider you configure.
Local-first agent originally from Block with CLI and desktop surfaces plus MCP extensions. Now stewarded by the Agentic AI Foundation under the aaif-goose GitHub org.
Self-host free; OpenHands Cloud is paid. Inference is BYOK or cloud-bundled.
Open software-agent platform with Agent Canvas UI, SDK, and self-hosted or cloud backends. Supports OpenAI-compatible providers when you configure them.
Free, Pro $20/month, Pro+ $60/month, and Ultra $200/month (Lite $8/month in selected countries), each with a token-based usage allowance plus optional on-demand usage.
Agentic IDE from ByteDance. Bundled models run through the product; you can also add models with your own API key, including custom OpenAI Chat Completions or Anthropic Messages URLs. Localhost endpoints are not confirmed, so they are not counted here.
Free with a Google account (60 requests/min, 1,000/day), or billed through a Gemini API key, Vertex AI, or a Code Assist license.
Open terminal agent for Gemini models with tools, MCP, and headless modes. Defaults to Google hosted APIs; treat arbitrary local OpenAI-compatible bases as unverified.
Free to start with BYOK for 10+ models; JetBrains AI subscriptions also cover Junie in the IDEs.
JetBrains coding agent available in IntelliJ-based IDEs, as a standalone CLI, and in GitHub and GitLab CI. Works with your own model keys or JetBrains AI; local endpoints are not confirmed.
Free (daily build credits), Pro $25/month and Business $50/month (100 monthly credits each), and Enterprise; credits cover building, hosting, and AI features.
Web agent that turns prompts into full-stack apps with hosted editing and deploy flows.
Xcode is free; ChatGPT and Claude integrations use those accounts. Custom and locally hosted chat providers are configurable in Intelligence settings.
Agent and chat surfaces built into Xcode, including Claude Agent access to Xcode build and test tools. Apple documents internet-hosted providers and locally hosted Chat Completions endpoints under Intelligence settings.
Closed to new signups since May 15, 2026; existing Pro subscriptions and IDE plugins are supported until April 30, 2027.
Retiring: AWS coding assistant with IDE plugins and a CLI. AWS directs users to Kiro; Q Developer in the AWS Management Console is unaffected. Agent features use Amazon hosted models.
Free (50 credits), Pro $20, Pro+ $40, Pro Max $100, and Power $200 per user per month, with add-on credits at $0.04.
AWS’s spec-driven agentic IDE and CLI and the replacement for Amazon Q Developer. Routes to Claude, GPT, and open-weight models through AWS; no bring-your-own-key or local endpoint path is documented.
Self-hosted and BYOK only; Plandex Cloud wound down in late 2025.
Dormant: terminal agent oriented around multi-file plans and accumulated diffs, configurable with OpenAI-compatible bases. The hosted cloud shut down and the repo has had no pushes since October 2025.
Free and open source; BYOK for any OpenAI-compatible provider.
Rust coding agent forked from OpenAI’s Codex and tuned for low-cost open models such as Kimi K3 and GLM 5.3. It can emulate other harnesses (Claude Code, Kimi Code, Qwen Code, SWE-agent), runs commands in native sandboxes, and works as an ACP agent. Replaces the older Python computer-control project.
Research agent is free; you pay the model provider.
Academic-origin agent framework for repository-level tasks with a defined agent-computer interface. Wire any compatible model endpoint you control. No tagged release since May 2025, though the repo is still updated.
Free during preview (3 workspaces; up to 30 for Google Developer Program members); no new workspaces, shutting down March 22, 2027.
Sunsetting: browser IDE evolved from Project IDX with Gemini-assisted app building and Firebase deploy paths. Google shuts it down on March 22, 2027 and points users to Google AI Studio or Antigravity.
Standalone Tabnine plans could not be confirmed after the Tricentis acquisition; tabnine.com redirects to Tricentis.
Long-running IDE completion and chat product with SaaS and private-deployment options, acquired by Tricentis in July 2026. Its context engine is being folded into Tricentis’s testing platform; confirm the standalone product’s status before recommending it.
Open editor fork with optional PearAI subscription and BYOK providers.
Inactive: open VS Code based editor that bundled agent chat via Continue and Roo Code/Cline forks and direct provider choice. No releases since May 2025.
Prime Agent
Claude Code
Codex CLI
Cursor
T3 Code
VS Code
GitHub Copilot
Pi
Hermes
Mercury
Oh My Pi
OpenClaw
OpenCode
Crush
Mistral Vibe
Letta Code
Kimi Code
Antigravity
Devin Desktop (formerly Windsurf)
Zed
Aider
Cline
Continue
v0
Warp
Amp
Factory
Roo Code
Kilo Code
Goose
OpenHands
Trae
Qwen Code
Junie
Replit Agent
Bolt
Lovable
Xcode Coding Intelligence
Amazon Q Developer
Plandex
Void
Augment
Open Interpreter
SWE-agent
Mentat
Tabnine
PearAI