Harness launched Autonomous Employee Brokers for software program supply: the platform for enterprises to construct and safely run AI brokers that deal with the work between writing code and delivery it to manufacturing.
Software program supply has moved via phases. First, individuals did the work by hand. Then they wrote scripts for particular person jobs like deployment. Most lately, they linked these jobs into automated pipelines that observe fastened directions, which is what Harness has run for big enterprises for years. Employee Brokers are the subsequent section. Each step within the pipeline can now run as a reasoning agent moderately than a hard and fast script, with the context, governance, sandboxing, and audit trails that enterprises have to belief brokers in manufacturing.
Dozens of Harness Managed Brokers can be found as we speak, and any crew can customise them or construct their very own. A brand new Harness Agent Market makes it simple to search out, use, and share them.
“AI now writes the code. Harness ships it,” mentioned Jyoti Bansal, Founder and CEO of Harness. “Autonomous Employee Brokers are how enterprises construct and safely run AI for every little thing after code: constructing, testing, securing, deploying, working. All of it runs on the identical pipelines that already ship our clients’ software program, inside their very own community boundary. The governance, the audit path, and the safety posture are already there. Employee Brokers inherit all of it from day one”.
The controls that hold brokers secure in manufacturing
Autonomous Employee Brokers run on infrastructure that the shopper controls. Code and information by no means depart the shopper’s community, and brokers are ruled by the identical controls enterprises already use for human deployments.
These controls make Autonomous Employee Brokers secure to run in manufacturing:
- Sandboxing: Brokers run in remoted containers with restricted file and community entry. An agent that produces a malicious command has nowhere to ship information.
- Scoped credentials: Every agent has its personal id and the particular set of permissions assigned to it, the identical approach an worker does. An agent can solely take the actions these permissions permit, regardless of who triggers it or what its immediate says.
- Coverage enforcement: The identical insurance policies that gate human deployments gate brokers. A coverage can hold brokers off non-approved fashions or out-of-production pipelines.
- Audit trails: Each agent motion is recorded underneath a definite AI id, with full provenance: what triggered the agent, what it did, and the end result.
- Value monitoring: Token spend is surfaced per agent, per pipeline, with price range caps that cease an agent when it hits its restrict.
- Chaining: Brokers compose into multi-step workflows, passing output from one to the subsequent.
Straightforward to construct, ruled by your insurance policies
Constructing an Autonomous Employee Agent makes use of the identical agent-file format that has develop into normal throughout the trade. Put it aside to a single file, commit it to your repository, and the agent is stay, ruled, and out there throughout your group. Groups that might moderately not write the file can use Harness AI to generate the agent for them. Both approach, the agent runs as a ruled pipeline step with the identical controls, audit path, and coverage enforcement as every little thing else.
As soon as it runs, the agent has your group’s full context. It causes utilizing the Harness Software program Supply Data Graph, a linked map of your companies, pipelines, deployments, infrastructure, incidents, and safety findings. An agent assessing a vulnerability is aware of which companies are affected and who owns them. A deployment agent is aware of which companies rely upon the one being deployed. The result’s a response constructed to your particular surroundings, not a generic repair that solely appears to be like proper.
Brokers additionally meet you the place you’re employed. By the Harness MCP Server, a developer in Cursor, Claude Code, or one other instrument can assign a activity to a Employee Agent and have it run in Harness, with the outcome returned to wherever it was triggered. Wherever an agent runs, it runs underneath your group’s insurance policies, ruled the identical approach as each different step in your pipeline.
Harness-built brokers, prepared to make use of as we speak
Harness has pre-built Autonomous Employee Brokers that deal with the repetitive, time-consuming work that slows groups down throughout the supply lifecycle. Listed below are a couple of of the brokers out there as we speak, with extra added repeatedly:
- Autofix reads construct logs, identifies the foundation reason behind a construct failure, commits a repair to the PR department, and re-triggers builds till it passes.
- Code Assessment evaluations PR diffs for code high quality, safety points, and take a look at protection.
- Code Protection identifies untested traces and generates exams to shut protection gaps.
- Characteristic Flag Cleanup detects stale flags and validates secure removing.
- Manifest Remediator analyzes failed Kubernetes deployments and fixes manifest points.
- IaCM Remediation fixes configuration drift, safety findings, and cloud value points by modifying infrastructure configurations.
The Harness Agent Market
The Harness Agent Market is a shared catalog the place Employee Brokers are revealed and reused throughout a company and the broader Harness neighborhood. Groups can undertake an present agent moderately than construct their very own, and contribute the brokers they construct again to the catalog.
It has three tiers:
- Harness Managed: Constructed, maintained, and SLA-backed by Harness.
- Harness Licensed: Constructed by companions, reviewed and authorized by the Harness engineering and safety groups.
- Neighborhood: Revealed by the broader Harness neighborhood. Organizations can use out-of-the-box insurance policies to manage which neighborhood brokers run in manufacturing.
Each agent within the Market will be forked. A crew can clone an present agent and modify the immediate, instruments, or triggers to suit their surroundings. The agent one crew builds to resolve an issue turns into the start line for the subsequent crew that hits the identical roadblock.
“We constructed RiskSentinel, a Harness Autonomous Employee Agent, to show that ruled AI can transfer past figuring out safety points to securely remediate them whereas sustaining enterprise controls, auditability, and compliance,” mentioned Ratna Devarapalli, Director IT, United Airways. “When constructing with Harness, what stood out most was how intuitive the expertise was — it enabled our crew to maneuver from an preliminary thought to a production-ready agent in simply 4 days, permitting us to deal with fixing an actual enterprise problem moderately than the underlying platform. That mixture of developer expertise and enterprise-ready capabilities is what is going to allow organizations to confidently scale AI throughout software program supply.”
Carry your individual mannequin
Autonomous Employee Brokers work with any LLM supplier. Join Anthropic, AWS Bedrock, or Google Vertex AI via present Harness connectors, and swap fashions per agent, per surroundings, or per pipeline with out rewriting the agent.
Availability
Autonomous Employee Brokers and the Harness Agent Market are actually usually out there to all Harness clients.