Amp and Warp vs Claude Code and Codex in 2026: free harnesses, $20 tiers, cloud runners and your own plan
Amp and Warp both went free this year if you bring your own Claude or ChatGPT plan, and both now sell $20 tiers, cloud machines and team billing on top. We read their pricing pages, docs and changelogs to show who pays for each run, what the cloud runners cost, and where the brief and design states should live while the agent works.

Two harnesses went free, and the bill moved somewhere else
In September Amp announced it was free to use if you bring your own compute and your own model subscription or key. A month later it added a Claude Code mode that runs on your Claude plan. Warp took a similar road: its Free plan accepts your own OpenAI, Anthropic or Google key, lets you sign in with ChatGPT, and gives Claude Code, Codex and OpenCode first-class support inside the terminal.
So the harness is no longer what you pay for. The money moved to three places: $20 paid tiers with included usage, cloud machines that run agents while your laptop is closed, and team plans with pooled spend. Meanwhile the ticket, the spec, the designed states and the launch visual still sit in six other tools, each with its own seat and its own AI meter.
That is the 2026 version of AI credit fatigue: not one big bill but many small allowances, and context that has to hop across MCP servers before an agent can start. This guide compares Amp and Warp with Claude Code and Codex from pricing pages, docs and changelogs checked on 22 November 2026, and shows where the shared context should live.
How Amp charges in November 2026
Amp, which Sourcegraph spun out as its own company in December 2025, has four ways to pay. They can be mixed inside one workspace.
- Hobby, free. Use your ChatGPT or Claude subscription, or your own API keys, with no Amp token fees or limits. Your model provider's charges and limits still apply. Self-hosted orbs and runners on your own machines are free and unlimited.
- Paid tiers, from $20 a month. The Individual tier is $20, and once signed in you can pick Megawatt 50 at $50 with 2.5 times Megawatt's included agent and orb usage. Gigawatt is the top tier and the only one with the largest orb size. Included usage resets monthly and does not roll over. Students and teachers get Megawatt for $10.
- Amp credits, pay as you go. No tier needed. For individuals and non-enterprise workspaces, Amp says it adds no markup to providers' API prices, so $2 of Anthropic usage costs $2 of credits. Web search and other tools also draw on credits, and purchased credits expire after twelve months.
- Orbs, billed per minute. An orb is a cloud computer where Amp works remotely. Hosted sizes run from a1.tiny at $0.08 an hour to a1.xxlarge at $1.32 and a1.3xlarge at $2.13. A paused orb costs nothing, and an orb pauses once the agent has been idle for 5 minutes and nobody has touched it for 20.
Teams cost nothing extra on top of the tiers and add pooled credits, SSO and shared keys. Since 10 October, picking Mode then Claude Code in a new thread runs the Claude Agent SDK on your linked Claude plan instead of Amp's own agent, and orbs, runners and thread sharing still work. Public threads ended on 2 June 2026; sharing inside a workspace remains.
How Warp charges in November 2026
Warp calls itself an agentic development environment: a terminal with its own agent, a CLI, and Oz cloud agents. Its plans are listed in dollars of agent usage at API rates.
- Free, $0. No included agent usage, but you can bring your own OpenAI, Anthropic or Google key, an OpenAI-compatible endpoint, or sign in with ChatGPT to use your Codex allowance. MCP is included.
- Build, $20 a month. $20 of agent usage at API rates, covering the Warp app, the Warp Agent CLI and Oz cloud agents. Bring your own key is allowed for individuals and organisations of up to ten people.
- Max, $200 a month. Ten times Build's included usage.
- Business, from $50 per user. $20 of usage per seat, SAML SSO, your own keys and custom inference endpoints, up to 25 seats. Enterprise adds inference in your own AWS Bedrock or Google Vertex account.
Extra usage can be bought on every plan, with auto-reload under a team spend cap, and stays valid for twelve months. Hosted Linux compute for cloud runs is metered per minute. On 17 November Warp added linked billing with the Claude Platform. Warp also runs Claude Code and Codex as they are, locally in its terminal and, in beta, inside Oz cloud environments, but handing a local Claude Code session to the cloud is not supported.
Amp, Warp, Claude Code and Codex side by side
All four read a repository, plan, edit files, run commands and connect to MCP servers, and all four read AGENTS.md (Claude Code reads CLAUDE.md). They differ in who sells the model access and what runs in the cloud.
| Agent | Plans | How a run is paid | Cloud runs and team features |
|---|---|---|---|
| Amp | Hobby free; Individual $20; Megawatt 50 $50; Gigawatt; Teams no extra charge | Your Claude or ChatGPT plan, your key, or Amp credits at provider price | Orbs $0.08 to $2.13 per hour, self-hosted free; workspace threads |
| Warp | Free; Build $20; Max $200; Business from $50 per user | Included usage at API rates, add-on usage, your key, or ChatGPT sign-in | Oz cloud agents, Claude Code and Codex in the cloud in beta |
| Claude Code | Claude Pro $20, Max $100, Ultra $200 | Your Claude plan's usage limits; API key billed per token | Claude Code on the web; CLAUDE.md and MCP |
| Codex | ChatGPT Plus, Pro and Business plans | Your ChatGPT plan's Codex allowance; API key billed separately | Codex cloud tasks; AGENTS.md and MCP in config.toml |
| Sydla workspace | $19 per user | No AI credits: drives your own Claude Code or Codex, zero token markup | Brief, criteria, UX Studio states and evidence in one MCP connection |
Read the table by the plan you already pay. Amp and Warp are now harnesses around the same Claude and GPT models you can reach directly. What they add is orchestration, cloud machines and team billing. Our bring your own coding agent guide explains the three ways to pay for a run, and the OpenCode comparison covers the open-source option.
The six-tool stack around any agent
Whichever harness runs the agent, it still needs a ticket, a brief, designed states, decisions and a launch visual. In a typical 2026 stack those sit in six tools, each with a seat price and an AI meter, and each one an MCP hop away from the run.
| Tool | Seat price | Bundled AI and its limit | Handoff to a coding agent |
|---|---|---|---|
| Jira Standard | $9.70 | Rovo: 25 credits per user per month, pooled; PAYG $0.01 per credit from 3 Dec 2026 | Atlassian MCP or a pasted ticket |
| Confluence Standard | $7.15 | Same Rovo pool, 25 credits per user | Atlassian MCP; enriched calls 1 to 10 credits |
| Notion Plus | $12.00 | Custom Agents on Business and Enterprise; extra credits $10 per 1,000 | Notion MCP server |
| Figma Professional Full | $20.00 | 3,000 AI credits per Full seat, no rollover | Figma MCP: 200 calls a day on Full or Dev |
| Miro Starter | $10.00 | 25 AI credits per license, pooled per team | Miro MCP: daily call limits by plan |
| Canva Business | $25.00 | Personal AI allowance per member | Outside the build loop |
| Sydla workspace | $19.00 | No AI credits: bring your own Claude Code or Codex | One MCP connection or a Sydla Desktop thread |
The six tools come to $83.85 per user per month before any overage, against $19 for Sydla, a saving of $64.85. Add a $20 Amp or Warp tier for the agent and the stack is $103.85 per person, with seven allowances to watch. The stack cost guide works through 5, 10 and 25 people, and the comparison overview covers each tool.
Technical walkthrough: your own agent on a shared brief and UX Studio states
Sydla does not add another harness or another meter. The context lives in one workspace, and the agent you already pay for reads it on your machine.
- Brief. Each ticket sits in an initiative with a versioned brief, numbered acceptance criteria and a done-means line. Every save is signed by its author, a person or an agent.
- UX Studio. Screens and their states, empty, loading, error and success, are designed next to the ticket and published as a handoff with tokens and a URL per state.
- Run. In Sydla Desktop you open the ticket, start a thread and choose claude or codex. Desktop drives the CLI you are already signed in to, so the model, the limits and the bill stay on your Claude or ChatGPT plan. Each thread has an access level, Read only, Edit files or Full access, with a Preview panel and a Changes view.
- Review. The agent records evidence against each criterion: test output, screenshots of each state and the diff. A person reviews it and decides the merge and release gates. An agent can request a gate but never approve one, and the record stays with the ticket rather than in a thread link.
Outside Desktop, the same context is one MCP connection away. For Claude Code, run claude mcp add with the HTTP transport and the sydla.com/mcp URL. For Codex, add an mcp_servers.sydla table to config.toml with the same URL and your key in the headers. Amp and Warp work too: both accept remote MCP servers, and Warp picks up MCP config from Claude Code and Codex, so an engineer who runs agents in an Amp orb or an Oz environment reads the same brief and states. Keep the first prompt short: read the work item, follow the criteria, record evidence. The brief to pull request walkthrough shows one feature end to end, and the Sydla Desktop setup guide covers the install.
Five checks before you add a harness
Amp and Warp are good tools. Before you roll one out to the whole team, check these five things.
- Which plan pays. If everyone already has a Claude or ChatGPT plan, Amp Hobby and Warp Free cost nothing extra. A $20 tier only pays off if you use its included usage or cloud runs.
- Cloud minutes. Amp orbs are billed per minute while running and Warp meters hosted compute per minute. Set spend caps before agents run unattended overnight.
- Team size limits. Warp's own-key option on Build and Max covers organisations of up to ten people; beyond that you need Business at $50 per seat. Amp Teams adds no seat fee.
- Claude plan terms. Amp's Claude Code mode uses the Claude Agent SDK on your linked plan, and Warp runs the real Claude Code CLI. Anthropic restricts third parties from routing requests through Pro, Max or Ultra plans, so check the current terms for your plan.
- Where context lives. A harness does not fix a brief spread across five tools. Decide where the ticket, criteria and states live before you choose where the agent runs.
When Amp or Warp is the right choice
Be fair to both. Amp is the strongest option if you want agents on cloud machines that keep working while your laptop sleeps, multiplayer threads, and model credits at provider price with no markup. Warp is the best terminal on this list, and it runs Claude Code, Codex and OpenCode side by side with notifications and a cloud handoff for its own agent.
They are a weaker fit if the whole team already pays for Claude or ChatGPT plans and only needs the agent on a laptop, or if the real problem is not the harness but a brief that lives in five places. Our Claude Code vs built-in AI agents guide, the Cline and Kilo Code comparison and the AI credit fatigue guide look at the same question from other sides.
Try it on your next ticket
Sydla is $19 per user per month with every feature included and no AI credits to buy. Your Claude Code or Codex plan does the coding, and the brief, the states and the evidence live with the ticket. See the pricing section, read how Sydla works, then start your workspace and keep your Claude Code and Codex accounts.
Frequently asked questions
Is Amp free in 2026?
Yes, on the Hobby tier. Since 13 September 2026 Amp is free when you bring your own ChatGPT or Claude subscription or your own API keys, with no Amp token fees, and self-hosted orbs and runners are free. You pay for Amp-hosted orbs per minute, Amp credits at the provider's API price, or a paid tier from $20 a month.
How much does Warp cost?
Warp Free is $0 with your own key or a ChatGPT sign-in. Build is $20 a month with $20 of agent usage at API rates, Max is $200 a month with ten times that, and Business starts at $50 per user with $20 of usage per seat, as of November 2026. Extra usage can be bought on every plan.
Can I use Claude Code or Codex inside Warp or Amp?
Yes. Warp runs the Claude Code, Codex and OpenCode CLIs in its terminal with first-class support, and can run Claude Code and Codex in Oz cloud environments in beta. Amp has a Claude Code mode since 10 October 2026 that uses the Claude Agent SDK on your linked Claude plan, and accepts a ChatGPT Plus or Pro subscription for OpenAI models.
What are Amp orbs and how much do they cost?
An orb is a cloud computer where Amp's agent works remotely. Hosted orbs are billed per minute while running, from $0.08 an hour for a1.tiny to $1.32 for a1.xxlarge and $2.13 for a1.3xlarge on Gigawatt. A paused orb costs nothing, and self-hosted orbs are free on Hobby, Individual and Teams.
Do I need Amp or Warp if I already pay for Claude Code or Codex?
Not to write code. Claude Code and Codex already run on your plan. Amp and Warp add cloud machines, multiplayer threads and a better terminal. What a small team usually lacks is one place for the brief, criteria and designed states. Sydla holds those for $19 per user and runs your own Claude Code or Codex with zero token markup.



