Atlassian 将 MCP 接口扩展至 220 个工具，并与 OpenAI 达成支出承诺

Atlassian 重建的 MCP 服务器现每天处理 1500 万次调用并暴露 220 多个工具，而新的支出承诺将 GPT-6 Astra 加入 Rovo 的多模型组合，同时不破坏平台模型无关的立场。

已发布 已更新 5 分钟阅读 ✓ AI 就绪 分享

TL;DR——三句话

  • Atlassian's rebuilt MCP server handles more than 15 million calls per day and exposes 220-plus tools across Jira, Confluence, Bitbucket, Loom and Goals (S1).
  • A new "spend commitment" runs from Atlassian to OpenAI; Rovo remains multi-model and Astra is explicitly not the default (S1).
  • Governance features include OAuth 2.1 + PKCE, per-user permission enforcement, tiered read/write/destructive actions, DLP via Atlassian Guard and org-wide audit logs (S1).

摘自 key_takeaways （真实字段，最多 3 条）。事实与分析分别标注。

Atlassian 将 MCP 接口扩展至 220 个工具，并与 OpenAI 达成支出承诺Editorial cover for a component sourcing brief. Chip motif; headline "Atlassian 将 MCP 接口扩展至 220 个工具，并与 OpenAI 达成支出承诺". Reported key numbers: 220 -plus. 1 cited source. Static cover (no animation).Atlassian 将 MCP 接口扩展至 220 个工具，并与 OpenAI达成支出承诺REPORTED KEY NUMBERS220 -plus1 sourceXAIPART
Cover image generated from data already shown on this page: the category motif, the headline, and reported key numbers. No decorative illustration; no third-party image.

您从 AI 回答跳转而来——这是可验证的原文

AI 摘要可能遗漏了其结论的依据。原始预测、受影响料号及所有引用均在本页，均可追溯至来源。

XAIPART 预测

本篇简报包含可证伪的预测。

跳转至完整预测及其证伪规则 ›
受影响的料号

此报道在目录中无匹配——我们选择不显示，而不是推断匹配。

了解判定方式 ›
需要采购料号？发送请求 → 询价

Reported by VentureBeat on . Summarized and analyzed by XAIPART Editorial.

What happened

Atlassian announced an expansion of its partnership with OpenAI, adding GPT-6 Astra and the GPT-5.6 series to its platform and Rovo AI, and connecting the Teamwork Graph to ChatGPT and Codex through a rebuilt Model Context Protocol server.S1 The arrangement includes a commercial commitment that an OpenAI representative described as "effectively a spend commitment" - Atlassian committing dollars to OpenAI models and technology, not the reverse.S1

Atlassian was explicit that GPT-6 Astra is not becoming Rovo's default model; Rovo routes requests dynamically across providers through an internal AI gateway that balances capability, speed and cost per task.S1 The company's own MCP announcement cites Anthropic's head of enterprise product, Scott White, calling Atlassian's MCP "one of the most-used enterprise MCP integrations on Claude," and Atlassian reports that its server's token-savings claims were measured on Claude models.S1

The rebuilt MCP server now handles more than 15 million calls per day, exposes 220-plus tools across Jira, Confluence, Bitbucket, Loom and Goals, and uses up to 25% fewer tokens on comparable Jira and Confluence work in internal testing.S1 Jira Service Management support is listed as coming soon.S1 Governance features include Atlassian-hosted endpoints secured with OAuth 2.1 and PKCE, per-user permission enforcement at query time, separated read, write and destructive risk tiers with confirmation gates on higher-risk steps, Data Loss Prevention via Atlassian Guard, domain and IP allowlists, per-scope controls and org-wide audit logging, including an option to disable writes entirely.S1

GPT-6 Astra pricing is listed at $10 per million input tokens and $50 per million output, with Rovo credit consumption scaling to the models and compute used per task.S1 OpenAI reports Astra reached the "Critical" cybersecurity threshold under its Preparedness Framework and scored 100% on one exploit benchmark, but the same safeguards can "slow, pause, or stop legitimate work," including defensive security tasks.S1 On data, Atlassian states only the data returned by an invoked MCP tool is sent to OpenAI, Astra supports Zero Data Retention for eligible API customers, enterprise access to Astra is off by default, and OpenAI is listed as a subprocessor that does not train on customer inputs or outputs.S1

Key numbers
MetricReported valueSource
MCP server daily callsMore than 15 million[S1]
MCP tool count exposed220-plus[S1]
Internal token reduction vs. prior serverUp to 25%[S1]
GPT-6 Astra input token price$10 per million tokens[S1]
GPT-6 Astra output token price$50 per million tokens[S1]
Spend-commitment counterpartyAtlassian to OpenAI (direction of payment)[S1]

Background timeline

  1. - Atlassian Intelligence launches, using OpenAI models from the start.S1
  2. - Original Atlassian-OpenAI partnership establishes the foundation later expanded.S1
  3. - Atlassian introduces its MCP integration for ChatGPT.S1
  4. - GPT-6 Astra ships as a standalone release.S1
  5. - Atlassian announces expanded OpenAI partnership and rebuilt MCP server, coinciding with Team '26 Europe.S1

XAIPART take

Technical impact

The MCP expansion turns Atlassian into a practical enterprise agent substrate: 220-plus tools across Jira, Confluence, Bitbucket, Loom and Goals, with per-user OAuth 2.1 + PKCE enforcement and tier-separated write/destructive actions that an admin can throttle or disable outright. For engineering teams already running Atlassian as a system of record, the 25% token reduction and the explicit DLP pipeline address the two biggest operational objections to long-running Rovo agents - cost and data egress - without forcing an all-in commitment to any single model family.XAIPART analysis

Market impact

Atlassian is paying OpenAI for priority model access while keeping Anthropic and other providers routed through the same gateway, which signals that model-portable enterprise infrastructure now has more commercial gravity than exclusive frontier-model deals. Buyers sourcing AI tooling should expect competing multi-model gateways from rival platforms, and any roadmap pinned solely to GPT-6 Astra will face the same benchmark ambiguity the source already flags on the Artificial Analysis Intelligence Index.XAIPART analysis

Action

Pilot the new MCP connection in a non-production Jira and Confluence tenant first, validate the OAuth 2.1 + PKCE enforcement and DLP labels against a representative prompt-injection scenario, and instrument Rovo credit usage per task before enabling Astra tier access. Lock the admin-console writes-disable switch and the per-scope read/write/search controls into your change-management baseline so that a model swap or an Astra safety pause cannot bypass your governance posture.XAIPART analysis

XAIPART Forecast

XAIPART expects that, within roughly two quarters of this announcement, at least one Atlassian-published Rovo reference architecture will demonstrate a third-party model performing greater than 50% of routed production traffic on the rebuilt MCP server, confirming the multi-model gateway as the durable layer rather than GPT-6 Astra. If, by , Atlassian has not disclosed a non-OpenAI model carrying more than 50% of Rovo traffic against this 15M-calls-per-day MCP baseline, this forecast is wrong.XAIPART analysis

Facts are as reported by the cited source. The design-in, sourcing and forward-looking statements above are XAIPART's analysis, not the source's claims. The contract size behind the "spend commitment," the precise scope of Astra Zero Data Retention eligibility, and whether the internal-Rovo model path matches the ChatGPT/Codex MCP retention terms are not disclosed by the source.

Corrections: none at publication. This brief will be updated with a timestamp if the source corrects its report.

Dated developments, extracted from this page's timeline (curated milestones = hollow markers).
Timeline of dated developmentsDated entries: 2023-04-01 2023 - Atlassian Intelligence laun…; 2023-01-01 - Original Atlassian-OpenAI partne…; 2025-12-01 2025 - Atlassian introduces its MC…; 2026-09-03 - GPT-6 Astra ships as a standalon…; 2026-10-06 - Atlassian announces expanded Ope….2023-04-012023 - Atlassian Intelligence laun…2023-01-01- Original Atlassian-OpenAI partne…2025-12-012025 - Atlassian introduces its MC…2026-09-03- GPT-6 Astra ships as a standalon…2026-10-06- Atlassian announces expanded Ope…

替代料与生命周期

已收录的替代料——仅为目录范围，不代表市场排名
料号关系生命周期依据
— 暂无收录替代料 待补充数据 — 需 match_type 后才可显示任何前代标签

“前代”和“替代”标签仅来自已记录的匹配类型、生命周期状态或来源文本。没有记录匹配时，不显示任何标签。我们自己的商品列表不构成市场参考。

受影响的料号

目录中无匹配。 此事件未映射到 XAIPART 当前上架的任何料号。我们不推断匹配，也不显示占位列表。

各维度成本压力

Source: FRED · Generated · reference output z and composite are algorithm outputs, not a fetch snapshot

Cost pressure by dimension — composite stateDiverging bars for three dimensions around a zero midpoint. Input Cost has no reading; Supply Demand z -1.4711, notable; Fx Energy z 0.2546, neutral. Composite index -9.1, band neutral.0投入成本no reading供需z −1.4711 · notable46 d汇率与能源z +0.2546 · neutral5 dcomposite −9.1 · neutral−100 · −40 · −15 · 0 · +15 · +40 · +100
Underlying values — copyable for verification.
DimensionFRED series IDUnitLagAs of
投入成本PCOPPUSDM, PALUMUSDM, PNICKUSDMUSD/mt52 d
供需PCU334413334413, IPG3344Sindex(1982=100) / index(2017=100)46 d
汇率与能源DEXCHUS, DCOILBRENTEU, DHHNGSPrate / USD/bbl / USD/MMBtu5 d

No free authoritative interface for: gold · silver · rare earth · polysilicon · lithium carbonate · market share — listed rather than approximated.

本文如何生成

过程证据——抓取、来源可靠性检查、模型与校验
本简报的审计记录
Source fetched VentureBeat · venturebeat.com · 8,210 chars
Source sanity (S1) ok — passed before any model call (no s1_ block recorded)
Model / prompt MiniMax-M3 · prompt Blog Agent - News Summary (v2.8-final) · 8,006 tokens · 33.4s · run completed
Gate verdict hold · 5/6 applicable checks passed · spec v2.9 · C10 ratio 0.010471 · C10 rounds 0
Time contract 7 <time datetime> values in this brief — 0 violations (+08:00 or date-only)

上述每一行均由流水线生成，而非人工撰写。

编辑评分

确定性校验结果——纯 PHP 实现，判定过程不涉及模型
校验评分明细 6 项适用检查， 5项通过： 83 %. Hold items: C15 (four views + no restating) — sent to human review. Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8343. 适用检查已通过 5 / 6 = 83% · 判定： hold Hold items: C15 (four views + no restating) — sent to human review Gate spec: v2.9 (S0 + S1 + C1–C20) · Source affinity (C20): 0.8343 是否发布由校验决定。独立审计评分仅供参考。

独立审计（仅供参考，不影响发布）： 51/100 · 等级 C — 详情

相关报道

发现错误？ 告诉我们哪里有问题 ——更正内容会与原文一同发布并记录在案。我们不会悄悄修改文章。

评论

成为第一个评论的人。

来源

  1. Atlassian deepens its OpenAI partnership while keeping its platform multi-model — https://venturebeat.com/orchestration/atlassian-deepens-its-openai-partnership-with-a-spend-commitment-but-its-platform-stays-firmly-multi-model · rss · 访问于

事实标注为 S1/S2，并可追溯至所链接的来源。标注为 分析 的内容是 XAIPART 自身的推理，与报道的事实分开呈现。来源链接会被访问核实，而非 nofollow ——署名是有意为之。