What shipped this week
Weekly product updates from the Fireweave team — new capabilities, improvements, and fixes across rollouts, guardrails, and control points.
-
Finish a release by hand, pick up setup & fw 3.1.0
You can Mark stable on a release that is not automatic, and a blocked ramp names why and offers Retry. Leave setup and come back — you land on the same step. fw 3.1.0 is on cli/latest.
-
A setup PR from Fireweave, Confirm what we found & Jobs
Fireweave reads your repositories, you confirm what it found, and it opens a setup PR per Managed repo — no local coding agent required. Finish setup and Jobs follow the run until you merge. The Fireweave SDKs are 3.0.0 on npm, PyPI, and crates.io, with a one-line start profile and Control points on the wire.
-
Opt-in GitHub analysis comments & a shorter first-project review
The advisory comment on a GitHub pull request is now a workspace opt-in — off by default — and creating the first project no longer asks a second acknowledgement.
-
Blast radius on the PR, a scored risk reading & one analysis comment
When change release analysis is on, the GitHub comment shows blast radius from your catalog, a scored release-risk reading you can open row by row, and a single comment that updates in place.
-
Repo-first onboarding, Watching & Release Risk Assessments
New workspaces walk a repo-first setup — GitHub, repositories, production branches — and land on a project home that watches those branches.
-
Manual ramp schedule edits & a hold on empty guardrails
Edit a live ramp’s future stages by hand — percents and holds — and a release with no guardrail metrics now parks for you instead of advancing as if everything were fine.
-
Searchable Guardrails, metric deep links & environment-bound preflight
Find a guardrail metric by typing in the catalog, jump from a rollout to that series in Datadog or OpenObserve, and preflight against the environment that is actually bound — so a configured provider no longer looks missing.
-
Datadog, Oodle & live guardrail checks
Connect Datadog or Oodle with an API key and use those streams as guardrail signals, pick metrics from a live catalog instead of typing names from memory, and pull a reading on demand with Check now — including a reason when a signal cannot be read.
-
Java Control points on Maven Central & a shorter initialise
Initialise now wires Java services to the ai.fireweave SDK on Maven Central, so a JVM surface can evaluate Fireweave Control points in production the same way TypeScript and Python already do.
-
Python Control points & optional initialise add-ons
Initialise now wires Python services to the fireweave package on PyPI, so a Python surface can evaluate Fireweave Control points in production the same way TypeScript already does. Observability and other add-ons stay offers — declining one that is off no longer parks the run.
-
AI Release Engineer, live ramp control & CLI 2.0
Choose how much of a release Fireweave runs for you, steer a live ramp — amend the schedule, pick a rollback target, end the soak — and install CLI and plugins 2.0, which keep rollout state on the server and let CI authenticate with an organization API key.
-
Control points for new projects & safer ramp defaults
New projects land on internally managed Fireweave Control points infrastructure — no external PostHog project to wire — and ramps refuse to start with their safety rails silently off.
-
Multi-project rollouts, org API keys & end-to-end tracing
Coordinate one rollout across multiple projects and surfaces, manage projects and environments programmatically with organization API keys, and get answers faster thanks to end-to-end tracing across the platform.
-
A new fireweave.ai, Google sign-in & sturdier ramps
The redesigned fireweave.ai is live, you can now sign in with Google, and a batch of fixes made in-flight ramps survive restarts, failures, and expired connections.
-
Tag-mapped environments, access-control hardening & a richer activity feed
Map git tags to environments so Fireweave knows where every change landed, benefit from a dedicated access-control hardening pass, and follow integration activity in a richer feed.
-
Instrument as you build, deploy-sdk on npm & staging-to-prod promotion
Coding agents can now wire flags, metrics, and deploy attestation while they write the feature, @fireweaveai/deploy-sdk installs straight from npm, and rollouts promote from staging to production.
-
Project rollout overview, OpenObserve connector & an org-wide dashboard
Every project gets a rollout overview with a stable flag lifecycle, OpenObserve joins the integration catalog, and a new organization dashboard shows every project at a glance.
-
Guardrail evidence on ramps & live console updates
Cohort ramps now advance on real stability and adoption metrics with the guardrail evidence to prove it, and the console updates in real time over a single org-scoped stream.
-
Organization invitations, manual ramp starts & smoother device sign-in
Invite teammates into your organization with a proper join flow, start a sealed rollout’s ramp on your own schedule, and sign in from the CLI without retyping codes.
-
Plugins across five coding agents, fw api & flags provisioned at seal
The Fireweave plugin now installs into five coding-agent hosts, the CLI talks to Fireweave Cloud directly through fw api, and sealing a rollout provisions its feature flags automatically.
-
The plugins marketplace, fw plugins CLI & leaner installs
Fireweave plugins for coding agents now install from a real marketplace via the fw plugins CLI, with atomic versioning and per-architecture bundles that download only what your machine needs.