{"slug": "meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans", "title": "Meta takes Muse Code out of beta and adds $5 monthly plans", "summary": "Meta took Muse Code out of beta on August 31st, adding cross-session messaging, workflows for coordinating teams of subagents, and monthly subscriptions starting at $5, 26 days after its beta launch. The release also puts a Muse Code SDK into developer preview, enabling developers to build around the coding agent's runtime. Meta's terminal coding agent, powered by the Muse Spark 1.2 model, now supports coordination across sessions and orchestrated subagent workflows, marking a shift from a model showcase to a paid agent platform.", "body_md": "# Meta takes Muse Code out of beta and adds $5 monthly plans\n\n**Meta's terminal coding agent adds cross-session messaging, multi-agent workflows and an SDK preview after a 26-day beta.**\n\nBy [Ryan Merket](/author/ryan-merket)\n· Published\n\nPrimary source: [Meta for Developers](https://x.com/MetaforDevs/status/2094502049912439068)\n\n## Why it matters\n\nMeta is turning Muse Code from a model showcase into a paid agent platform. Cross-session messaging, workflows and an SDK make coordination, rather than code completion alone, the product bet.\n\nMeta took Muse Code out of beta on August 31st, adding cross-session messaging, workflows for coordinating teams of subagents and monthly subscriptions starting at $5.\n\nThe release comes 26 days after Meta introduced Muse Code as a beta product. In [a post on X](https://x.com/MetaforDevs/status/2094502049912439068), Meta said the monthly plans were beginning to roll out on August 31st. Meta also put a Muse Code SDK into developer preview, giving developers an early route to build around the coding agent's runtime.\n\nMuse Code runs in the terminal and uses Meta's [Muse Spark 1.2](/models/meta/muse-spark-1.2) model to plan changes, edit code and validate its work across large repositories. Meta launched the model and agent together on August 5th, describing Muse Code as a harness built around persistent background agents rather than disposable workers created for a single request.\n\n### From persistent subagents to coordinated sessions\n\nThe most consequential addition is communication between sessions. Meta's original beta architecture assigned persistent subagents to a main session, with those agents deciding when to report information back to the primary agent. The August 31st release extends coordination across sessions, allowing separate Muse Code jobs to exchange messages while they work.\n\nThat matters when engineering work has been divided across repositories, services or parallel investigations. A session changing an API can pass information to another session updating the client, for example, without requiring a developer to copy context between terminal windows. Meta has not claimed that the sessions share complete transcripts or working directories, so the feature should be understood as messaging rather than automatic synchronization.\n\nWorkflows add another layer of coordination. Meta says developers can use them to orchestrate teams of subagents, turning Muse Code's existing parallel workers into repeatable processes. The announcement does not specify how many agents a workflow can operate or what controls developers have over branching, retries and spending.\n\nMeta had already built Muse Code around a local append-only event log that records model calls, tool runs, approvals and edits. According to Meta, that design lets an interrupted session resume from the point where it stopped. Muse Code's original skills included commands for producing an approval-gated plan, challenging that plan and working toward a pinned goal.\n\nThe public [Muse Code cookbook](https://github.com/meta-models/meta-model-cookbook/blob/main/04_muse_code/README.md) also documents recipes for replaying sessions in continuous integration, isolating subagents in separate Git worktrees, scheduling recurring work and running side conversations outside the main transcript.\n\n### Meta puts a subscription in front of the agent\n\nThe $5 starting price gives Muse Code a low-cost subscription path as Meta moves the product beyond its short beta. Meta described the new plans as a rollout, meaning availability may reach accounts in stages.\n\nThe pricing decision is part of the product strategy. Coding agents increasingly consume unpredictable amounts of inference as they read repositories, run tools and delegate work to parallel agents. A monthly plan gives individual developers a simpler entry point than calculating token consumption for each job, although the starting price alone does not show how much agent usage the plan includes.\n\nMuse Code enters a market where orchestration has become a central product feature. OpenAI's Codex app is designed to supervise multiple agents running in parallel, and OpenAI has expanded Codex from terminal and cloud coding tasks into longer-running automations and broader computer work. ([openai.com](https://openai.com/index/introducing-the-codex-app/?utm_source=openai)) Meta's response is centered on a terminal harness whose model and agent environment were trained together.\n\nMeta said Muse Spark 1.2 was co-trained with Muse Code using agent trajectories, goal-oriented recipes, context compaction and subagent behavior. Meta also reported testing the pairing on GPU-kernel optimization runs lasting as long as 24 hours and spanning more than 1,000 tool calls. Those performance examples remain Meta's internal evaluations rather than independent evidence of how Muse Code compares with other coding agents on everyday repositories.\n\nThe SDK preview pushes the product beyond a standalone terminal application. Developers can begin evaluating Muse Code as a component for internal engineering systems, automated maintenance jobs and specialized development tools. Combined with workflows and cross-session communication, the release shows Meta packaging Muse Code as an orchestration layer for agent teams, with the $5 plan designed to get more developers running those teams.", "url": "https://wpnews.pro/news/meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans", "canonical_source": "https://runtimewire.com/article/meta-muse-code-out-of-beta-monthly-plans", "published_at": "2026-08-31 20:39:17+00:00", "updated_at": "2026-08-31 20:53:26.093726+00:00", "lang": "en", "topics": ["ai-products", "ai-agents", "developer-tools", "artificial-intelligence"], "entities": ["Meta", "Muse Code", "Muse Spark 1.2", "Meta for Developers"], "alternates": {"html": "https://wpnews.pro/news/meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans", "markdown": "https://wpnews.pro/news/meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans.md", "text": "https://wpnews.pro/news/meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans.txt", "jsonld": "https://wpnews.pro/news/meta-takes-muse-code-out-of-beta-and-adds-5-monthly-plans.jsonld"}}