Loading today’s stories
Build1 distinct publisher3 min readUpdated
The Agents SDK now carries durable execution, sandboxed code execution and a durable filesystem. Cloudflare's argument is that harnesses cannot own those problems, which makes them a platform decision.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Cloudflare has pushed durable execution, dynamic code execution, a durable filesystem and dynamic workflows into the Cloudflare Agents SDK, where any harness built on the SDK can use them [7]. The framing matters more than the feature list: Cloudflare argues these problems cannot be solved inside a harness at all, because they are tied to state, storage and compute, and therefore to the platform the agent runs on [6].
That claim comes out of the company's own harness work. Cloudflare built Project Think as its first-party agent harness [4], and says that running agents in production with customers surfaced the same distributed systems problems every time: how an interrupted agent resumes without losing context or re-spending tokens, how it runs untrusted code securely, and how it gets the tools the model was trained for [5]. The published version of the argument is wrapped in the usual calendar rhetoric, that 2026 is the year agent harnesses go to production [2], which can be ignored without losing the engineering point.
Cloudflare draws the result as three layers: a framework on top providing project structure, conventions, integrations, CLI and developer experience; a harness in the middle running the agentic loop that calls tools, reads results and manages context; and the runtime underneath supplying compute, state and storage [8]. Flue, an open-source framework from the team behind Astro, is the first to build on that bottom layer [9]. It shipped 1.0 Beta this week on the Pi harness, the same harness OpenClaw uses [10].
Flue's approach is declarative: rather than scripting an orchestration loop, you describe the context an agent needs, its model, skills, sandbox and instructions [11]. Cloudflare's worked example is a triage agent that intercepts a bug report, reproduces it in a sandbox and diagnoses it in under 25 lines [12]. Around that sit pre-configured Channels for Slack, GitHub, Linear and Discord that absorb event verification and dispatch boilerplate [13], React hooks in @flue/react that stream agent state, tool execution and live messages into a frontend [14], and a CLI that generates a Markdown blueprint your own coding agent can edit when you run something like `flue add channel slack` [15]. Crash tolerance is handled by Durable Streams, an append-only log of every prompt, tool response and model choice, so a host crash, an LLM provider timeout or a restart does not wipe the turn [16].
The consequence for anyone shipping this is coupling, and it runs in the direction most teams do not plan for. If resumability, sandboxing and durable filesystem semantics live in the runtime, then picking the runtime narrows the set of harnesses that can actually use them. Cloudflare names five harnesses as production-grade in its own post, Codex, Claude Code, OpenCode, Pi and Project Think [3], and only two of the five, Pi and Project Think, appear in the stack it drew [17]. The primitives are advertised as available to any harness on the Agents SDK [7], which is a real offer and also a boundary condition: the harness has to be on the Agents SDK.
Worth pricing before committing: whether a harness you already run can adopt these primitives without a rewrite, whether durable state written through the SDK is portable off Cloudflare, and how the durable log behaves on replay when a tool call has already had an external side effect.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Cloudflare is bringing durable execution, dynamic code execution, a durable filesystem and dynamic workflows to the Cloudflare Agents SDK as a base layer, now available to any harness building on the Agents SDK.
Flue shipped 1.0 Beta this week, built on the Pi harness, the same harness that OpenClaw is built on.
Flue's model is declarative: instead of scripting what the agent does or writing an orchestration loop, the developer defines the context an agent needs, including its model, skills, sandbox and instructions.
Cloudflare's example is a triage agent that intercepts a bug report, reproduces it in a sandbox and diagnoses the issue in under 25 lines.
Flue addresses host crashes, LLM provider API timeouts and unexpected restarts via Durable Streams, adding each event in the execution history to an append-only log and treating every prompt, tool response and model choice as an unchangeable ledger so agent state is never volatile.
Cloudflare published a post announcing it is bringing more agent harnesses and frameworks to Cloudflare, starting with Flue.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single vendor primary source, specific but unverified
All material comes from one Cloudflare blog post. It is unusually concrete for a vendor announcement -- named SDK methods, named packages, a described durability mechanism and a three-layer architecture -- which raises its evidentiary value above marketing boilerplate. But there is no second publisher, no benchmark, no failure-injection or recovery measurement, and no independent user report, so every capability and maturity claim rests on the party selling the platform.
One beta framework and vendor-disclosed harness reuse
Observable adoption is narrow: a 1.0 Beta framework described as the first to build on the Agents SDK, the SDK primitives themselves shipping, and a vendor statement that the Pi harness also underlies OpenClaw plus Cloudflare's own Project Think. There are no deployment counts, named customers, download or usage figures, and no evidence that the three third-party harnesses named in the post's opening (Codex, Claude Code, OpenCode) build on the SDK.
Production framing outruns beta-stage, single-source evidence
The post opens by declaring 2026 the year harnesses go to production and agents load-bearing infrastructure, and titles a section 'Designed for production, not just prototypes', while the flagship artifact is a 1.0 Beta and the durability claims ('state is never volatile') are asserted rather than measured. The architectural argument that durability belongs to the platform is coherent and internally consistent, which keeps the gap moderate rather than severe, but the naming of five harnesses against a stack containing only two -- one of them Cloudflare's own -- widens the stated-versus-shown distance.
First-party platform announcement with direct consumption upside
The sole source is the vendor whose product is the subject. Cloudflare's argument -- that resume, sandboxing and state cannot be owned by the harness and are therefore a platform decision -- routes buying pressure to its own Agents SDK, Durable Objects and Workers consumption, and the post also promotes Cloudflare's own first-party harness, Project Think. The named ecosystem partner, Flue, is described as 'our new open-source framework', further aligning the framing with the publisher's commercial interest. Nothing in the supplied material offsets this with independent or adversarial commentary.
High confidence in what was said, low in whether it holds up
Confidence is high that the announcement, architecture description and feature list are accurately captured -- the primary source is authoritative for its own release. Confidence is low on the load-bearing questions a reader actually needs: whether the durability guarantees hold under real failure, whether the ecosystem extends past Pi-based frameworks, and what any of it costs. One publisher, one beta artifact and zero independent measurement cap the score near the midpoint.
build
Flue 2 bets that agents are a rendering problem, not an orchestration one1 distinct publisher
build
Claude Code now outruns Copilot roughly two to one in JetBrains' survey of 15,000 developers1 distinct publisher
invest
Vercel's eve bets the agent plumbing layer consolidates the way Next.js did1 distinct publisher
product
A 2x LLM bill is not a bug report: token spend is an observability problem1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.