BTC $64,958.99 +0.55%
ETH $1,819.49 +0.33%
BNB $590.29 -0.05%
XRP $0.9606 +1.05%
SOL $71.36 +0.59%
TRX $0.3070 +0.66%
DOGE $0.0594 +1.52%
ADA $0.1535 +0.41%
BCH $208.72 +0.42%
LINK $9.21 +0.35%
HYPE $67.19 +1.26%
AAVE $107.60 +0.47%
SUI $0.7226 +0.58%
XLM $0.1378 -0.09%
ZEC $892.76 -0.12%
AAPL $353.47 -0.85%
AMZN $258.64 -1.81%
GOOGL $286.39 -1.74%
MSFT $521.91 +0.44%
META $627.78 -0.83%
NVDA $156.47 -2.21%
TSLA $379.20 -2.03%
SNDK $822.82 -3.67%
INTC $55.74 -3.23%
SPCX $137.12 -1.01%
MU $730.57 -3.48%
AMD $509.75 -2.09%
BTC $64,958.99 +0.55%
ETH $1,819.49 +0.33%
BNB $590.29 -0.05%
XRP $0.9606 +1.05%
SOL $71.36 +0.59%
TRX $0.3070 +0.66%
DOGE $0.0594 +1.52%
ADA $0.1535 +0.41%
BCH $208.72 +0.42%
LINK $9.21 +0.35%
HYPE $67.19 +1.26%
AAVE $107.60 +0.47%
SUI $0.7226 +0.58%
XLM $0.1378 -0.09%
ZEC $892.76 -0.12%
AAPL $353.47 -0.85%
AMZN $258.64 -1.81%
GOOGL $286.39 -1.74%
MSFT $521.91 +0.44%
META $627.78 -0.83%
NVDA $156.47 -2.21%
TSLA $379.20 -2.03%
SNDK $822.82 -3.67%
INTC $55.74 -3.23%
SPCX $137.12 -1.01%
MU $730.57 -3.48%
AMD $509.75 -2.09%

agent

All
Article
Flash

Anthropic opens self-hosted private testing for Agent, with execution records retained by the enterprise

Anthropic announced the opening of Claude Managed Agents for self-hosted private testing. Enterprises can deploy the AI Agent's operating framework, execution sandbox, access credentials, and session history on their own cloud servers or internal data centers, without having to rely on Anthropic for hosting. In May of this year, Anthropic allowed enterprises to self-host the sandbox, enabling AI to run code on internal servers, but the core systems responsible for scheduling Agents, calling models, and saving session records were still managed by Anthropic.The new version also places these control systems within the enterprise intranet. When enterprises allow Agents to automatically fix code, test software, or troubleshoot system failures, they can restrict the servers that can be accessed, manage API keys, and use OpenTelemetry to log and review each operation step.The Claude model itself still runs in the cloud. When the Agent infers, the relevant prompts, code, and tool results still need to be sent to the model service provider. Enterprises can choose to call Claude through Anthropic, Amazon, Google, or Microsoft's cloud services, or they can route all model requests through their own servers to uniformly control access permissions, calling channels, and costs. This feature is currently only available for private testing to approved applicants, and enterprises need to contact the Anthropic customer team to obtain deployment documents.

The pass rate and cost of Alma by independent developer Yetone are better than Claude Code

The locally developed AI Agent Alma, led by independent developer Yetone, ranked third in the latest round of cross-evaluation initiated by the AI Agent infrastructure company Runta, both in task success rate and cost of success, outperforming tools like Claude Code and OpenCode. The current leaderboard snapshot shows that Alma has a task success rate of 72.7%, just behind Codex CLI at 77.9% and Pi at 72.8%, while Claude Code stands at 70.0%.Alma costs approximately $0.80 on average for each successfully completed task, compared to Claude Code's $0.94, making it about 16% cheaper. This FrontierHarness v1.1 comparison evaluates the performance of the same large model under different Agent execution frameworks, covering 21 sets of Harness configurations, 5 large models, and 62 tasks, with each task repeated 4 times, planning to run a total of 26,040 times, which is 72 times the scale of the previous version. All tools operate in a unified cloud environment, comparing task success rates, token costs, and execution speeds.Alma emphasizes local operation and long-term memory, capable of reading and modifying code, executing terminal commands, searching the web, and supporting Skills and MCP, while also allowing for switching underlying models. Recently, Alma integrated with Herdr, allowing programming Agents like Claude Code, Codex, and Pi to continue operating within Herdr, and enabling users to view execution processes, respond to questions, and approve operations in a centralized chat interface. Users can also remotely manage multiple Agents on their computers via iPhone. The evaluation is still ongoing, and rankings may change; the models being tested are limited to open-source models like GLM, DeepSeek, Kimi, and Muse, and do not cover closed-source models from the GPT and Claude series.

first_img Space and Time, supported by Microsoft, launched an enterprise platform that allows AI agents to verify private data

According to a report by Decrypt, the blockchain data project Space and Time, supported by Microsoft, announced the launch of the Space and Time Enterprise platform, which allows enterprises to use private data for AI tools without relinquishing the underlying records. The platform can prove that computations based on private data are correct, enabling AI applications, auditors, and other relevant parties to verify results without accessing the database itself.The first batch of partners includes on-chain asset management firm Maple Finance, Indonesian automotive manufacturer and distributor Indomobil, Oracle software consulting company Trenez, and The Data Foundation, a nonprofit organization focused on government data based in Washington, D.C.Nate Holiday, CEO of the original developer of Space and Time, MakeInfinite Labs, stated that enterprises want AI agents to utilize this information to execute workflows, but the agents do not need unrestricted access to the records behind every computation.He mentioned that granting unrestricted data access to AI tools would create new exposure channels. The company is collaborating with clients in construction, engineering, and manufacturing to enable agents, reporting tools, and compliance teams to work based on the same set of business records and verify the computations they rely on.The platform is based on Space and Time's patented technology, Proof of SQL, which is used to prove that database queries are computed correctly and is currently compatible with Google Cloud's BigQuery. Enterprises can run it on their own servers without needing to operate a blockchain, purchase dedicated hardware, or rely on external proof services.With GPU acceleration, supported queries on millions of rows of data can be completed in one second and generate proofs. This technology can only prove that computations are correctly executed on given records and cannot determine whether the records themselves are accurate. Space and Time completed a $20 million strategic financing round led by Microsoft's venture fund M12 in 2022, and Grayscale previously launched a trust product for the SXT token.
app_icon
ChainCatcher Building the Web3 world with innovations.