Skip to content
agora

ADR-0017: Agora is source-available under BSL 1.1 (superseding FSL-1.1-MIT)

Context

Section titled “Context”

The repository was first licensed under the Functional Source License 1.1, MIT Future License (FSL-1.1-MIT) — committed 2026-05-22 per the agora MVP spec. FSL is source-available: it forbids any “Competing Use” and converts each version to the permissive MIT license two years after release.

The offload V1 delivery spec (docs/superpowers/specs/2026-05-29-agora-offload-v1-design.md, decision V1-D2) revisited the license as part of productizing the offload stack and chose the Business Source License 1.1 (BUSL-1.1 SPDX) instead. The two specs therefore disagreed, and the offload-launch wave is where §8 of the V1 spec turns license choice into shipped packaging — forcing the question to be settled in one direction.

The decision turns on adoption friction for the users V1 targets: self-hosters and security/compliance-conscious teams running agents against their own repos, credentials, and (self-hosted) regulated data.

  • FSL’s “Competing Use” restriction is broad and somewhat fuzzy. A company with an internal agent platform has to stop and ask “is this a competing use?” — and that hesitation is exactly where an evaluation stalls.
  • BSL’s restriction is narrow and explicit: you author an Additional Use Grant, and ours forbids only offering Agora (or a derivative) as a hosted or managed orchestration / agent-dispatch service. A team evaluating Agora to run their own agents reads that and immediately knows they’re clear.
  • BSL is the recognized incumbent (MariaDB, CockroachDB, Sentry-then, HashiCorp) — lower legal-review friction than the newer FSL.
  • BSL’s “no hosted service” line maps onto the architecture already built: the §10.6 client/service privilege split is the commercial boundary, and the future hosted multi-tenant control plane is the service side.

Nothing is published yet (private: true, version: 0.0.0), so this is a clean swap with no relicensing of shipped artifacts.

Decision

Section titled “Decision”

Agora is source-available under the Business Source License 1.1, superseding the earlier FSL-1.1-MIT choice. Parameters (LICENSE, with the canonical BSL terms incorporated by reference to https://mariadb.com/bsl11/ to avoid reproducing the copyrighted template verbatim):

  • Licensor: Quarry Systems.
  • Licensed Work: Agora (this repository).
  • Additional Use Grant: all use permitted except offering Agora, or a derivative, to third parties as a hosted or managed orchestration / agent-dispatch service.
  • Change Date: 2030-06-01 (four years from first publish; the date in LICENSE is authoritative and advances with major releases).
  • Change License: Apache License, Version 2.0.

"license": "BUSL-1.1" is set on the root and every workspace package.json; a plain-language summary lives in LICENSING.md. All public copy says “source-available (BSL)” — never “open source” (BSL is not OSI-approved).

Consequences

Section titled “Consequences”

What becomes easier:

  • Self-host adoption: the one restriction is explicit and narrow, so an evaluating team (and its lawyers) can clear it quickly. The self-host compliance story (regulated data never leaves the customer’s account) is unobstructed because production self-host use is plainly permitted.
  • Commercial clarity: “self-host everything; we monetize only the hosted control plane” is a clean, honest story that matches the §10.6 split.
  • Familiarity: reviewers have seen BSL before; less friction than FSL.

What becomes harder:

  • Less aggressive competitive protection than FSL: BSL (with our grant) lets others build competing tooling, forbidding only reselling Agora itself as a hosted service. That breadth was FSL’s edge — traded away deliberately, since the same breadth chills the legitimate early adopters V1 needs first.
  • Longer closed horizon: 4 years → Apache-2.0, versus FSL’s 2 years → MIT.

Trade-offs:

  • We accept weaker anti-fork protection and a longer conversion horizon in exchange for the lowest adoption friction and an architecture-aligned commercial boundary. For an unshipped, early-stage project optimizing for getting real users, reducing evaluation friction beats maximizing protection.
  • This ADR supersedes the FSL-1.1-MIT choice (which predated the ADR series and was recorded only in the MVP spec + the 8e33e4e/63e34b1 license commits), not an existing ADR.