CLI
Explore from your terminal
Browse talks, speakers, agendas, and engineering tools. Use JSON output in your own scripts.
>
Agent Engineering HQThe home of the agent engineering discipline. Technical events, media, and a conference for engineers building production AI agents.
Install AgentEng
Next conferenceAgent Engineering San Francisco Coming April 2027/san-francisco Previous conferenceAgentEng London 2026 16 October 2026 · Canary Wharf/london-2026
Looking for our meetups? San Francisco events.
$ agenteng events --previous
Our last event filled AWS Builder Loft in San Francisco to capacity.
1,100+
registrations
300+
in the room
4
technical talks
>cat README.md
What is Agent Engineering?
// our definition
Agent Engineering is the discipline of designing, building, evaluating, and operating AI agents. It covers prompt, context, harness, inference, memory, evaluation, and code.
Agent Engineering HQ brings that discipline together in conferences and meetups. Our first conference, Agent Engineering London 2026, was fully booked, with nine technical talks on a single track. The next conference is Agent Engineering San Francisco, coming April 2027, and between conferences we run technical meetups in San Francisco and London.
It is for people who ship agent systems: engineers, researchers, and technical leads who work on the stack behind the model.
Grown from practitioner communities in San Francisco and London, including nearly 5,000 builders in London Agentic AI.
$ agenteng disciplines
Prompt
Context
Harness
Eval
Memory
Inference
Loop
Agentic
Code
Protocol
Graph
Search
Factory
Decision
Open-source agent tooling
Agent Engineering,
in practice.
AgentEng isn't just a conference. It's open-source agent infrastructure and tooling for agent engineering.
Come for the events, stay to tinker. Play with MCP, CLI, A2A, and ACP agents, learn the infrastructure, and do real agent engineering. We're building AgentEng alongside the events, so you can build along with us.
agenteng / build with it
Find tools for your next build
agenteng tools --discipline memoryExplore the memory tool directory. Try other disciplines such as eval, harness, and inference.
CLI setup and examplesMCP
Connect your coding agent
Give your agent access to conference context and the tool directory through local or remote MCP.
A2A
Build an agent integration
Discover the AgentEng agent and query it through A2A. Reuse the same request contract over HTTP.
ACPExperimental
Turn ideas into runnable demos
Launch a coding agent with talk or tool context attached, and ask it to build something in your workspace.
Learn the tools. Tinker with the infrastructure. Build something real.
- 1Explore the tools
- 2Connect your agent
- 3Try an idea
- 4Build something real
>
From Prompt Engineering to Agent Engineering
AI system engineering keeps specializing. In practice, these disciplines converge in one place: Agent Engineering.
discipline::01
Prompt Engineering
discipline::02
Context Engineering
discipline::03
Harness Engineering
discipline::04
Eval Engineering
discipline::05
Memory Engineering
discipline::06
Inference Engineering
discipline::07
Loop Engineering
discipline::08
Agentic Engineering
discipline::09
Code Engineering
discipline::10
Protocol Engineering
discipline::11
Graph Engineering
discipline::12
Search Engineering
discipline::13
Factory Engineering
discipline::14
Decision Engineering
discipline::∞
*.Engineering
convergence::active
Agent Engineering
Prompts, context, harnesses, evals, memory, inference, loops, protocols, graphs, search, and code converge in systems engineering.
$ all_roads --lead-to agent_engineering
Prompt Engineering
Context Engineering
Harness Engineering
Eval Engineering
Memory Engineering
Inference Engineering
Loop Engineering
Agentic Engineering
Code Engineering
Protocol Engineering
Graph Engineering
Search Engineering
Factory Engineering
Decision Engineering
discipline::∞
... *.Engineering
convergence::result
Agent Engineering
Agentic systems require more than prompts alone. They depend on prompts, context, harnesses, evals, memory, inference, loops, protocols, graphs, search, and code working together as one engineering discipline.
>
Two Cities, One Discipline
Agent Engineering HQ runs San Francisco. London events run with London Agentic AI. Both hold the same standard of technical depth.
agent-engineering-hq
San Francisco
Home base of Agent Engineering HQ. Practitioner forums on harness engineering, memory engineering, inference engineering, and more.
london-agentic-ai
London
Events with London Agentic AI, the UK's biggest Agentic AI community, including the AgentEng London Conference 2026.
$ agenteng events --sf
Meetups and partners
Join a meetup in San Francisco or London, or help with the next edition if it suits you.
Meetups
Next up in San Francisco: Code Engineering: From Coding Agents to Software Factories, 27 October 2026, 6:00 PM. Follow the calendars to hear about the next one.
Partners, optional
Support is optional and comes with no obligation. Help with food, drinks or recording if it suits you.
Partner with us>
The Agent Engineering Manifesto
manifesto.md
01
Agents are not features. They are long-running systems with their own development lifecycle, and they need their own engineering discipline.
02
Agentic systems behave differently from traditional software. A response can look successful and still fail the task. Behaviour drifts. Trajectories matter as much as outputs. These properties demand a dedicated Agent Development Lifecycle (ADLC), distinct from the traditional software development lifecycle.
03
The ADLC covers building, testing, deploying, and monitoring agents in production, with iteration and governance running as continuous layers throughout. Production signals feed back into evaluation. Traces drive improvement. Governance keeps the system trustworthy.
04
Agent Engineering is the practice of designing, building, evaluating, and operating AI agents across that lifecycle. It applies to every category of agent, from coding and research agents to customer-facing, domain-specific, and autonomous multi-agent systems.
05
AgentEng exists to define and advance this discipline. We bring together the practitioners building the tools, frameworks, and supporting infrastructure that make every phase of the ADLC work in production.
>
The Agent Engineering Mindset
Non-Determinism
Agent systems do not behave deterministically, so reliability has to be engineered around variance in output rather than assumed from the code path.
Intelligent Specification
Agent behaviour follows the specification it is given. Precise task definitions and planning up front produce more predictable results at runtime.
Every Input is an Edge Case
Users can submit any input, so conventional test coverage does not hold. Behaviour is better understood through evaluation against production traffic.
Resource Allocation
Compute, budget and human review are finite resources, and allocating them across agent runs is a core operational concern at scale.
The Reviewer Framework
Engineering work shifts from writing code towards designing the reviewers that check it, using validation loops and automated pull request gates.
Agent Networking
Agents working together need defined communication patterns, with parallel and sequential workflows arranged to avoid conflicting or duplicated work.
>
Core Engineering Themes
The engineering disciplines that form the backbone of production-grade Agentic AI.
disciplines.index · 14 modules
live
01
✨
Prompt Engineering
Writing effective prompts, system instructions, and prompt patterns. Covers structured outputs, few-shot design, prompt optimization, and LLM-native communication.
02
🧭
Context Engineering
Shaping what an agent sees at runtime. Covers context construction, compression, grounding, retrieval, and MCP (Model Context Protocol).
03
🧰
Harness Engineering
Defines the execution environment around an agent. Wires models to tools, policies, sandboxes, and execution constraints.
04
🧪
Eval Engineering
Measuring agent behavior in non-deterministic systems. Covers evaluation frameworks, behavioral testing, LLM-as-judge, and reliability benchmarks.
05
🧠
Memory Engineering
How agents store, retrieve, and evolve state over time. Covers short-term vs long-term memory, retrieval strategies, and personalization.
06
⚡
Inference Engineering
Serving models efficiently in production. Covers GPU/hardware selection, inference engines (vLLM, SGLang, TensorRT-LLM), latency budgeting, batching, and cost optimization.
07
🔁
Loop Engineering
Designing the agent's run loop: the iterate, observe, and act cycle. Covers control flow, stopping conditions, retries, self-correction, and long-running autonomous loops.
08
🤖
Agentic Engineering
Automating the entire software development lifecycle and engineering with AI agents.
09
💻
Code Engineering
Improving the coding capabilities of models and agents. Covers agentic coding, coding agents, code generation quality, and AI-assisted software development.
10
🔌
Protocol Engineering
Standing up the protocol layer agents talk over. Covers MCP, ACP, A2A, A2UI, transport and auth choices, versioning, and cross-vendor interoperability.
11
🕸️
Graph Engineering
Building agents with graphs. Covers graph databases, knowledge graphs, graph retrieval, and modelling agent state and execution as graphs.
12
🔎
Search Engineering
Agentic search techniques and the retrieval layer underneath them. Covers query planning, ranking, hybrid search, and improving agent behaviour through better search.
13
🏭
Factory Engineering
Designing and building software factories, where agents take work from spec to shipped code. Covers pipelines, review gates, orchestration of many agents, and keeping the output maintainable.
14
⚖️
Decision Engineering
Engineering the systems that make decisions. Covers decision models, decision-system design, planning under uncertainty, and how agents choose and justify an action.
>14 disciplines loaded_
>
Multi-Agent Conference Planning
Watch our agent swarm coordinate in real-time to create the most technically rigorous Agentic AI conference.
LIVE
Orchestrator
Venue
Speakers
Sponsors
Audience
Content
agent-swarm://agenteng-2026
COMMUNITY
Active
Builder Network
THEMES
6
Core Disciplines
PLATINUM
Invite Only
By invitation
GLOBAL
London
Then San Francisco
A Technical Conference for Agent Engineers
Production patterns, failure analysis, and scaling strategies, presented by the engineers who build these systems.
Held 16 October 2026
>
The Ecosystem
From frameworks and infrastructure to models, tools, and platforms.
agentic-coding
Agentic Coding
- $coding-agents
- $ai-pair-programming
- $code-review-for-agents
- $spec-driven-development
- $code-security-for-agents
active
frameworks
Agent & Tooling
- $agent-frameworks
- $orchestration-platforms
- $sdks-and-developer-tooling
- $multi-agent-systems
active
models
Models and Foundation
- $frontier-model-providers
- $developer-platforms
- $applied-ai-tooling
- $foundation-models
active
dev-tools
Agent Dev Tools & Frameworks
- $development-frameworks
- $apis-and-sdks
- $testing-tools
- $evaluation-frameworks
- $mcp-tools
active
agentops
AgentOps & Traceability
- $llm-trace-&-observability
- $opentelemetry-for-agents
- $model-serving-infrastructure
- $inference-engines
- $agent-sandboxes
active
enterprise
Enterprise & Security
- $iam-for-agents
- $tool-use-guardrails
- $permission-frameworks
- $banking-workflows
- $compliance-automation
active
>
Further Reading on Agent Engineering
Agent Engineering surfaced independently across practitioner communities, platforms, and research.
- 1March 2025Why Agent Engineering
- 2December 2025Agent Engineering: A New Discipline
- 3December 2025Reliable Agent Engineering
- 4June 2025Agent Engineering as a Core Practice
AgentEng brings these conversations together, focused on practice over promotion.
// organisers
Organised by Superagentic AI
Built on community, shaped with practitioners, speakers, and collaborators.
Superagentic AI
Advancing the Agent Engineering discipline and practice
leadership@agenteng ~
$agenteng whoami --chair
[ Shashi Jagtap ]
Conference Chair & Curator · Ex-Apple Engineer (6 years) · Organizer, London Agentic AI & Agent Engineering HQ
Building the bridge from Developer Experience to Agent Experience. Transatlantic connections between London and Silicon Valley, with programming shaped by a broader technical community.