Continue.dev alternatives after the Cursor acquisition

Continue.dev is finished as a product: continue.dev is now a one-page farewell that says "Continue was acquired by Cursor", the GitHub README says the repository is "no longer actively maintained and is read-only for all users", and the hosted platform behind it no longer exists. The clients still install and still run against your own endpoint, so you are not forced to move today — but the login, the hub and the account are gone, and the documentation site does not say so. The live alternatives worth moving to are Cline and opencode; the one to avoid is Roo Code, which is archived.

One disambiguation before anything else, because this query is polluted. This page is about Continue Dev, Inc. — the coding agent at continue.dev with a CLI, a VS Code extension and a JetBrains plugin. It is not the continue loop-control keyword, not Visual Studio's "Edit and Continue" debugger feature, and not continue.com, which is an unrelated longevity research organisation on the shorter domain.

Continue.dev acquired by Cursor: what is actually verifiable

Three dates can be checked directly, and the announcement date cannot. The continue.dev site bundle hard-codes a sunset anchor of 2026-06-15T00:00:00Z with three phases — early, mid, final — currently pinned to mid. The final VS Code release was published June 19, 2026. The New Stack covered the acquisition with a JSON-LD publication date of June 22, 2026. For the announcement itself that article hedges — it says Continue updated its homepage "around June 16" — and no first-party notice carries a date at all, so this page treats June 16 as reported rather than established.

The README also claims "a final 2.0.0 release of the VS Code extension, CLI, and JetBrains plugin", including removal of anonymous telemetry and authentication. The registries disagree with two thirds of that, and the registries are what you install from.

All four checked September 19, 2026. At the v2.0.0-vscode tag, extensions/cli/package.json still reads version 0.0.0-dev, and there is no CLI 2.x git tag — so the honest statement is that the CLI's latest published npm version is 1.5.47, not that a CLI 2.0.0 never existed anywhere. The telemetry and authentication removals are a README assertion; no shipped build was inspected here. The README additionally recommends the CLI over the JetBrains plugin.

Continue.dev pricing: there is no price, and no recoverable old one

Continue's FAQ states that "Recurring billing has been disabled, and you will not incur any new subscription charges from Continue going forward", that billing questions go to support@continue.dev, and that "Any unused paid credits remaining in your account will be refunded" — excluding used, promotional, trial, complimentary, gift and support-issued credits, payment processing fees and taxes. That is what Continue says; whether any refund was actually issued is not something this page can verify.

continue.dev/pricing returns HTTP 404, and the hosts that carried the paid platform are either gone from DNS or no longer answering, so no current plan price exists to quote and no historical one could be recovered from a live source. Treat any "Continue was $X per month" figure in an alternatives listicle as unverified. The clients themselves are Apache-2.0 and free; your only ongoing cost is the model API bill on whatever endpoint you point them at.

What still works, what is gone, and the trap in the docs

This is the part worth reading slowly, because docs.continue.dev returns HTTP 200, carries no acquisition or sunset banner on the pages a new user lands on, and still lists "A Continue account, or an Anthropic API key" under Requirements alongside cn login, which it says "opens your browser to authenticate with Continue". Half that sentence still works: the Anthropic-key path is local. The account half points at a platform that is gone.

Those dead-host readings come from DNS lookups, connection timeouts and HTTP status checks, plus Continue's shipped documentation. Nothing was installed or executed, so this page does not describe how the CLI behaves when it reaches a host that does not answer.

Keeping the Continue CLI and VS Code extension on your own endpoint

Continue's OpenAI provider page documents provider: openai as the protocol rather than the vendor, with an apiBase override for compatible endpoints. Continue's CLI configuration page gives the resolution order — the --config flag first, then the saved config, then the default ~/.continue/config.yaml; the source joins that directory onto the home directory, so on Windows it lands at %USERPROFILE%\.continue\config.yaml. Back the file up before you touch it — cp ~/.continue/config.yaml ~/.continue/config.yaml.bak — because restoring that one file is the entire rollback.

name: continue-after-sunset
version: 0.0.1
schema: v1
models:
  - name: Agent model
    provider: openai            # the protocol, not the vendor
    model: claude-sonnet-4-6
    apiBase: https://api.kunavo.com/v1
    apiKey: sk-kn-...
    roles: [chat, edit, apply]
    capabilities: [tool_use]    # documented as required for Agent mode

  - name: Autocomplete model
    provider: openai
    model: claude-haiku-4-5
    apiBase: https://api.kunavo.com/v1
    apiKey: sk-kn-...
    roles: [autocomplete]

Two source-level details worth knowing, both read at the released v2.0.0-vscode tag rather than from the docs. First, ignore the usual trailing-slash folklore: core/llm/index.ts appends a slash to apiBase when one is missing, and core/llm/llms/OpenAI.ts then resolves each endpoint relatively against it, so /v1 and /v1/ behave identically. What you must not omit is the /v1 itself — Continue's own 404 handler prints a hint saying exactly that.

Second, the real footgun. With provider: openai, Continue silently routes to POST on your base URL plus /responses instead of /chat/completions when the model id matches /^o[0-9]+/ or /gpt-[5-9]/ and you have not set useResponsesApi: false; for those same ids it also swaps max_tokens for max_completion_tokens. A gateway that only implements chat completions fails on exactly those ids; Continue's own provider page documents the escape hatch, useResponsesApi: false, and describes the trigger more narrowly than the code does — the page says "o-series and gpt-5 models" while the regex reaches gpt-5 through gpt-9. Kunavo's API serves /v1/chat/completions, /v1/responses and /v1/messages (verified in this repository, September 19, 2026), so that branch has a counterparty either way — a route-existence check, not a compatibility test. The id in the block above, claude-sonnet-4-6, matches neither pattern, so it takes the chat-completions path. Setting provider: anthropic with an apiBase also resolves to your base URL plus /messages in the source, but that override is undocumented on Continue's Anthropic page and untested here — do not treat it as a supported path.

Continue.dev alternatives: which one fits which user

Repository figures from the GitHub API, checked September 19, 2026. Two acquisitions appear on this table and they are not the same kind of event: Continue's product was shut down, its backend deleted and its billing stopped, while Kilo is still shipping under a new name on a new domain. Star counts and push dates are cadence evidence, not quality judgements, and none of these clients was runtime-tested here.

Continue.dev migration: what moves, and what you rewrite

Neither Cline nor opencode reads .continue/rules. Cline's rules documentation lists .clinerules/, .cursorrules, .windsurfrules and AGENTS.md with no Continue entry (checked September 18, 2026); opencode's rules page documents AGENTS.md (checked September 19, 2026). This is a comparison of documented file shapes, not a migration anyone ran — no real Continue config was moved into either client and executed.

A safe order: keep your Continue install exactly as it is and copy ~/.continue somewhere before you start; stand the new client up beside it and give it one bounded task; only then remove anything. Because the Continue clients no longer depend on a backend, leaving the old install in place costs nothing and is your rollback.

Best model for Continue.dev, and a worked cost estimate

Continue splits work across roles, and the autocomplete role fires far more often than chat. That is where the money goes, and it is the single most useful thing to get right in the config above. The figures below are illustrative token arithmetic, not measured task costs and not a bill ceiling. Assume one agent session of 200,000 uncached input tokens and 12,000 output tokens, and a day of autocomplete at 400 requests averaging 2,000 input and 50 output tokens each. Rates are live Kunavo catalog prices per million tokens.

Under these assumptions the autocomplete column alone moves from $0.360 on Claude Haiku 4.5 to $1.080 on Claude Sonnet 4.6 for the same keystrokes — which is why the role split matters more than the headline rate. Scale both columns by your own sessions and editing days before treating them as a budget; the autocomplete request shape in particular is an assumption, not a measurement.

Kunavo's catalog amount is a billing floor rather than a cap: when the upstream reports its charge, the bill is the greater of catalog cost and upstream cost times the applicable markup. Cache charges and external tools sit outside this example. The minimum top-up is $10 in prepaid credit — a funding minimum, not a task fee or a subscription. See billing details.

Where to go next

If you are staying on the final Continue build for now, the Continue integration guide has the config block in full. If you are moving, start with Cline setup or the OpenCode integration, and best API for Cline for the provider comparison. Kunavo publishes setup guides for these clients; a published configuration reference is not a compatibility test, and none of them has been runtime-tested here against Kunavo's endpoint. Create a Kunavo account when you are ready to fund a key, then run one bounded task and read the charge your account recorded for it.

Related reading: OpenAI-compatible API for the protocol itself, Cline pricing and OpenCode pricing for two clients with the same free-client, metered-API split, and Cursor alternatives if the acquirer is where you were considering going.

FAQ

Checked September 19, 2026: continue.dev and its site bundle, the GitHub README and releases API, the npm registry, the VS Code Marketplace and Open VSX listings, the JetBrains plugin API, docs.continue.dev, DNS and HTTP status for hub.continue.dev and api.continue.dev, and GitHub repository metadata for Cline, opencode, Kilo, Aider, Zed and Roo Code. Continue's request routing, and the hub and login removals, were read from source at the released v2.0.0-vscode tag; what the CLI on npm still reaches for was read at the v1.5.45 tag. Cline's rules documentation was checked September 18, 2026, and Anaconda's Kilo acquisition post on September 21, 2026. Nothing was installed, launched or called: there are no runtime or compatibility tests on this page. Kunavo token rates come from the live catalog, and every dollar figure here is illustrative token arithmetic.