Continue.dev pricing

About

Continue.dev builds an open-source AI coding agent and IDE extension (VS Code and JetBrains) that lets developers create, run, and share custom AI agents wired to their own models and tools. The core extension is free and open-source; the company monetizes a hosted layer — model credits, agent management, and team governance — through three plans.

Pricing splits along two axes: token-metered consumption for individuals and per-seat subscriptions for teams. The entry Starter plan carries no monthly fee and bills frontier-model usage at $3 per million tokens (input and output) pay-as-you-go. Team is $20 per seat per month and bundles $10 in model credits per seat plus centralized agent controls and SSO. Company is custom-quoted for enterprises needing SAML/OIDC SSO, BYOK, and contractual SLAs.

For the most current information on Continue.dev’s pricing and market position, visit Continue.dev.

Pricing summary : How Continue.dev’s pricing model works

Continue.dev runs a hybrid model layered on top of a free open-source extension. Individuals pay only for what they consume — frontier-model tokens at $3 per million (input and output combined) with no monthly fee. Teams move to a per-seat subscription at $20/seat/month that includes $10 of model credits per seat, so the bill blends a fixed seat fee with metered model usage. Enterprises negotiate a custom commitment.

Billing dimensions in play:

• Model tokens — $3 per million tokens (input + output), pay-as-you-go on Starter; also drawn against included/purchased credits on Team.
• Seats — $20 per seat per month on Team.
• Bundled credits — $10 in model credits included per Team seat each month.
• Custom commitment — Company tier adds invoicing, BYOK, and SLA under a negotiated contract.

What makes this different: The free, open-source core means Continue monetizes hosted convenience (model credits + agent management) rather than the IDE itself. Starter is unusual in having literally zero subscription floor — a developer can use paid frontier models purely on metered tokens.

Pricing by product

Continue Platform (all plans)

The open-source Continue IDE extension (VS Code / JetBrains) is free; the table above covers the hosted Continue platform. Frontier-model usage is metered at $3 per million tokens on Starter and drawn against included or purchased credits on Team.

Sales motions across products: PLG / self-serve for Starter and Team (sign up and pay by token or seat); sales-led for Company (custom quote, invoicing, SLA).

Hidden costs : What Continue.dev users actually pay

Continue’s headline numbers look cheap, but the real bill is driven by model token consumption, not the seat fee. The $20/seat Team price is essentially a governance fee; the variable cost is whatever your developers burn against frontier models at $3 per million tokens. Agentic coding (multi-step plans, large-context edits, repeated reads) consumes tokens far faster than chat-only autocomplete, so the metered line is where bills grow.

The seat fee is half credit, half license. Each Team seat’s $20 includes $10 in model credits, so a developer who stays under $10 of model usage effectively pays a $10 net license. A developer running agents all day blows past the $10 credit and pays $3/M on everything above it — the same rate as Starter.

Worked example — 5-developer team, moderate agent use:

The same five developers on Starter (no seat fee) would pay only for tokens: at ~12M tokens/dev/month that is ~60M tokens × $3/M = $180/mo, but with no shared agents, SSO, or admin controls. Team’s premium buys governance, not cheaper tokens.

The BYOK escape hatch. Cost-sensitive teams can sidestep Continue’s $3/M metered credits entirely: the open-source extension lets you bring your own provider API keys (or run a local model via Ollama) so you pay your LLM provider directly and Continue’s hosted credits become optional. On the Company tier, BYOK is a first-class feature; on Starter/Team it is the open-source fallback. This makes Continue’s effective token cost a ceiling, not a floor.

Want to estimate your own Continue.dev bill? Use the Continue.dev pricing calculator to model seat fees plus token overage against your team’s usage.

Pricing evolution : Continue.dev pricing history and changes

Continue spent its first ~18 months as a pure free, open-source project with no paid plan at all. Monetization arrived in stages: a tiered model announced with Continue Hub in February 2025, a live pricing page in January 2026, and a quiet removal of the free individual tier by March 2026.

Cadence

Tracked range: 2023–present. Historical inflection points sourced from Wayback Machine snapshots of continue.dev/pricing (2026-01-30, 2026-03-05, 2026-06-03) plus TechCrunch and Hacker News coverage.

Notable changes

• 2023-07-26 — Launched as a free, open-source coding autopilot (Apache 2.0). No paid plan. Show HN hit 298 points / 103 comments.
• 2025-02-26 — Shipped 1.0 and launched Continue Hub with a $3M seed round (Heavybit). Introduced a tiered model: free Solo, Teams (multiplayer + admin), Enterprise (security, audit log, credential management). Source: TechCrunch.
• 2026-01-30 — First live pricing page. Three tiers: Solo (Free), Team ($20/seat/mo, $10 credits), Company (custom). The path was a 404 as recently as 2025-12-29.
• 2026-03-05 — Entry plan renamed Solo → Starter and switched from Free to $3 per million tokens, pay-as-you-go (input and output). This removed the free hosted individual tier; the open-source extension remained free.

The free-tier removal in detail

The most consequential change is the quietest. In the January 2026 pricing page, the individual plan was literally called Solo and priced Free — “for individuals getting started,” with BYO API keys or the option to buy credits. By the March 2026 snapshot the same column had been renamed Starter (“Try agents and get started”) and re-priced to $3 / million tokens, input and output, pay as you go. Nothing on the Team or Company tiers moved.

This is a textbook usage-based pivot: rather than gate the free tier with seat limits or feature caps, Continue simply attached a meter to the hosted convenience layer. A developer who brings their own keys (or runs Ollama locally through the open-source extension) still pays nothing to Continue; a developer who wants Continue’s managed frontier-model access now pays per token from the first request. The free software never went away — the free hosted usage did.

What’s unique : Continue.dev’s distinctive pricing mechanics

1. The product is free; only the convenience is metered. Unlike GitHub Copilot or Cursor — where the editor integration is the paid product — Continue’s IDE extension is Apache-2.0 open source and free forever. Continue monetizes a separate hosted layer (Continue Hub: model credits, shared private agents, governance). This means a determined team can run Continue at $0 to Continue by bringing their own model keys, and the paid plans compete on convenience and governance, not on access to the software.

2. A zero-floor metered entry plan. The Starter plan has no monthly minimum — not “$0 for limited use” but literally no subscription line at all. You pay $3 per million tokens, input and output combined, and nothing else. Most “free or cheap entry” coding tools either cap free usage hard or charge a flat $10–20/mo floor; Continue’s entry plan is pure pay-as-you-go on managed frontier models.

3. The seat fee doubles as a usage wallet. Team’s $20/seat isn’t a pure license — it bundles $10 of model credits, so half of each seat fee is pre-paid usage. The bill then blends a fixed seat component with metered token draw against (and beyond) that credit. It’s a hybrid where the “subscription” and “usage” halves are deliberately fused inside a single seat price, which keeps the headline simple while still scaling cost with consumption. See how AI companies blend seats and usage for the broader pattern.

Strengths & weaknesses

Billing UX : Continue.dev billing controls and transparency

• Buy credits for frontier models — Starter lets users purchase model credits directly to run paid frontier models on a pay-as-you-go basis.
• Per-seat included credits — Each Team seat includes $10 in model credits per month, visible against metered usage.
• Agent usage controls — Team admins can control which agents the team can use and manage/share private agents centrally.
• SSO login — Gmail/GitHub SSO on Team; Custom SSO via SAML or OIDC on Company.
• Bring your own API keys (BYOK) — Company tier lets enterprises route model calls through their own provider keys.
• Commitment & invoicing — Company tier adds commitment-based invoicing and an SLA in place of self-serve card billing.

Strategic wins : Why Continue.dev’s pricing decisions worked

1. Open source as distribution, hosted convenience as revenue

Continue gave away the editor integration (Apache 2.0) to win developer trust and distribution — 32k+ GitHub stars and 2.4M+ VS Code installs by early 2026 — then monetized a separate hosted layer (Continue Hub) that the free users gradually want. This is the classic open-core split done cleanly: the thing developers evaluate is free and un-gated, so adoption is frictionless, and the paid product (managed credits, shared agents, governance) sells itself once a team forms around the free tool. See usage-based pricing strategy for why metering the convenience layer rather than the software lowers buyer resistance.

2. A token meter instead of a free-tier teardown

When Continue removed its free individual tier in early 2026, it didn’t claw back features or impose seat caps — it simply attached a $3/M-token meter to the hosted plan. Buyers who object can still pay $0 by bringing their own keys. That avoids the backlash that usually follows free-tier removals, because nothing was taken away — the free path just moved from “managed credits” to “your own keys.” This mirrors the broader shift from per-user licenses to consumption that’s reshaping developer-tool pricing.

3. Fusing the seat fee and the usage wallet

Team’s $20/seat bundling $10 of credits is a quiet design win: it gives buyers a predictable, comparable seat price (easy to budget against Copilot’s per-seat fee) while still letting cost scale with real usage. The seat fee anchors the value perception; the embedded credit absorbs light usage; overage captures heavy users at marginal cost. Picking the seat as the headline metric and tokens as the scaling metric is a deliberate value-metric choice — see choosing the right usage metric.

Areas to improve : Gaps in Continue.dev’s pricing approach

1. Token-metered agents invite bill shock

Agentic coding is token-hungry and bursty — a single large refactor or multi-step plan can consume millions of tokens. With $3/M billing and only $10 of included credit per seat, a heavy day can quietly produce a bill many times the seat fee, and there’s no obvious spend cap or budget alert surfaced on the public pricing page. Continue should ship visible per-seat/per-team spend caps and forecasting to avoid the bill-shock pattern that plagues consumption-priced AI tools.

2. The free-tier removal isn’t explained on the pricing page

The Solo→Starter change quietly converted “Free” to “$3/M tokens” with no migration note, changelog entry, or explanation visible on continue.dev/pricing. Developers who remember a free Solo plan now arrive to find a meter and no context. A short “what changed and why” note (and a clearer signpost that BYOK/self-host remains free) would reduce confusion and reinforce the open-source promise.

3. A two-point self-serve ladder with a large gap to Company

There are only two self-serve prices — $3/M and $20/seat — and then a cliff to a sales-led custom quote for anything involving SSO, audit, or BYOK governance. Mid-sized teams that need SAML or audit logging but aren’t ready for a negotiated contract have nowhere to land. A published “Business” price point between Team and Company (fixed SSO + spend controls at a transparent rate) would smooth the ladder.

Key takeaways

1. The editor is free; the hosted layer is the product. Continue’s IDE extension is Apache-2.0 open source — you only pay for the managed Continue Hub layer (model credits, shared agents, governance), and BYOK/self-host stays free.
2. Starter has no monthly floor. It’s pure pay-as-you-go at $3 per million tokens (input and output combined), with no subscription minimum — unusual among coding tools that gate or charge a flat entry fee.
3. Team’s $20/seat is half license, half usage wallet. Each seat bundles $10 of model credits; usage beyond that is billed at the same $3/M as Starter, so Team buys governance, not cheaper tokens.
4. The free hosted individual tier was removed in early 2026. The “Solo (Free)” plan visible in January 2026 became “Starter ($3/M tokens)” by March 2026 — a quiet conversion of free managed usage into a meter.
5. For the AI-coding category, open-core + a metered convenience layer is a viable counter-model to Copilot/Cursor’s pay-for-the-editor approach: give away the integration, monetize the managed inference and team controls.

UBP implications

1. Meter the convenience, not the software. Continue removed its free tier by attaching a token meter to the hosted layer rather than gating features — because a free self-host/BYOK path still exists, the change avoided the backlash a true free-tier teardown usually triggers. If you have an open-source core, this is the gentlest way to introduce usage pricing.
2. Fuse the seat and the meter inside one price. Bundling $10 of usage credit into a $20 seat gives buyers a comparable, budgetable headline (seat) while cost still scales with consumption (tokens). Practitioners adding usage to a seat-based product can fuse the two into a single line to keep the headline simple.
3. A zero-floor metered entry plan lowers the trial barrier. Starter’s “no monthly fee, $3/M tokens” removes the commitment of a subscription minimum, letting price-sensitive individuals try managed inference at marginal cost — a useful pattern for converting a free-software user base into paying usage customers.

Sources

• Continue.dev pricing page (accessed 2026-06-08)
• Continue.dev official website (accessed 2026-06-08)
• Continue documentation (accessed 2026-06-08)
• Continue blog (accessed 2026-06-08)
• Continue on GitHub (Apache 2.0) (accessed 2026-06-08)

Bottom line

Continue.dev is the open-source counter-model in the AI-coding space: give away the editor integration (Apache 2.0, 2.4M+ installs), monetize the hosted convenience layer. Its pricing is genuinely simple — a zero-floor Starter at $3/M tokens, a $20/seat Team that bundles $10 of credits, and a custom Company tier — but the simplicity hides a 2026 free-tier removal (Solo Free → Starter $3/M) and a token meter that can make heavy agent use cost far more than the seat fee. The honest pitch: if you bring your own keys or self-host, Continue can cost you nothing; if you want managed frontier-model access and team governance, you pay per token from the first request.

Want to compare Continue.dev against other AI-coding and developer-tools companies? Browse the pricing blueprint, or model your own bill with the Continue.dev pricing calculator.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Three-tier plan structure captured

Jun 2026

continue.dev/pricing shows Starter at $3/million tokens (PAYG, no monthly fee), Team at $20/seat/month with $10 included credits per seat, and custom-priced Company tier with SAML/OIDC SSO, BYOK, and SLA. Verified against live capture.

Free Solo tier replaced by metered $3/M-token Starter

Mar 2026

Wayback snapshot shows the entry plan renamed from 'Solo' (Free) to 'Starter' priced at $3 per million tokens, input and output, pay-as-you-go. Team ($20/seat/mo, $10 credits) and Company (custom) unchanged. This removed the free individual tier on the hosted platform; the open-source extension stayed free. Source: web.archive.org snapshot 20260305 of continue.dev/pricing.

First paid pricing page goes live (Solo / Team / Company)

Jan 2026

continue.dev/pricing first returns HTTP 200 (the 2025-12-29 capture was a 404). Entry plan 'Solo' is Free (BYO API keys or buy credits), Team is $20/seat/mo with $10 monthly credits per seat, Company is custom with SSO/BYOK/SLA. Source: web.archive.org snapshot 20260130 of continue.dev/pricing.

Continue 1.0 + Continue Hub launch with tiered Solo/Teams/Enterprise model

Feb 2025

Alongside its 1.0 release and a $3M seed round (led by Heavybit), Continue announced Continue Hub and a tiered model: a free Solo tier for individuals, a Teams tier with multiplayer/admin controls, and an Enterprise tier with security, credential management and audit logging. Source: TechCrunch, 2025-02-26.

Launched as free open-source coding autopilot (Show HN, YC S23)

Jul 2023

Continue debuted as a free, open-source AI coding autopilot (Apache 2.0) with no paid plan. Show HN reached 298 points / 103 comments. Y Combinator S23 batch; ~$2.1M seed followed post-YC. Monetization (Hub, credits) came later. Source: Hacker News, 2023-07-26.