Humans
01People still discover products the human way, and that discovery still has to land.
- Websites
- Search
- Social
- Events
- Communities
- Content
Reads a page
Software adoption engineering
Your software should be built for developers and agents.
We engineer the ecosystem that makes software discoverable, understandable, buildable, and agent-ready.
AgentRel helps software companies grow through developer relations, technical content, documentation, developer experience, AI discoverability, MCP, agent integrations, developer education, communities, and emerging protocols.
Starts with a scored audit — no engagement required.
One capability · create_invoice · four consumers
NeedsA person needs a reason to care. Positioning, proof, a next step.
The position
We engineer software adoption. Not attention, not impressions — adoption.
- Not a digital marketing agency
- Not an SEO agency
- Not a generic AI consultancy
- Not a DevRel staffing firm
The layer between software companies, developers, AI assistants, and autonomous agents.
AgentRelSoftware productAdoption
DevelopersAPI · SDKDocs · Examples
AI assistantsLLM contextGEO · AEO
AgentsMCP · A2AA2UI · x402
The shift
Software marketing changed.
Traditional marketing is optimized for attention. Modern software is adopted, and adoption is a chain of technical events — not a funnel of impressions.
Optimized for attentionTraditional
- 01Ads
- 02SEO
- 03Social
- 04Landing pages
- 05Lead generation
- 06Sales
Optimized for adoptionAgentRel
- 01Discover
- 02Understand
- 03Try
- 04Build
- 05Integrate
- 06Automate
- 07Transact
- 08Advocate
A developer doesn’t buy software because they saw an ad. They adopt it because they can understand it, test it, integrate it, and trust it.Developer reality
An AI agent doesn’t read your marketing funnel. It needs structured capabilities, tools, context, permissions, and reliable interfaces.Agent reality
The AgentRel model
Discover. Understand. Build. Integrate. Automate. Transact. Advocate.
Seven stages, in order. Each one has its own artifacts, its own failure mode, and its own measurement. Skipping one is where adoption stalls.
Discover
Make the product visible where technical buyers actually look — and where machines now answer on their behalf.
What we build here
- Developers and technical search
- Search engines and answer engines
- AI systems and coding assistants
- Communities and ecosystems
Outcome: the product appears in the answer, not just the index.
Four consumption surfaces
Four kinds of user. One product. Four completely different reading formats.
Most companies build for the first surface, tolerate the second, and have never considered the last two. All four now decide whether the product gets adopted.
Developers
02The people who decide whether the product survives the evaluation.
- Documentation
- APIs
- SDKs
- Tutorials
- GitHub
- Postman
- Examples
- Playgrounds
Runs a call
AI assistants
03Increasingly the first thing a developer asks about your product — and the thing that writes the integration.
- Structured documentation
- Product knowledge
- APIs
- Examples
- LLM context
- Technical references
ChatGPT · Claude · Gemini · Copilot · Cursor
Autonomous agents
04Software that selects your product, calls it, handles the failure, and pays for it.
- Tools
- APIs
- MCP
- A2A
- Structured interfaces
- Authentication
- Permissions
- Payments
- Observability
Executes an action
Surfaces 03 and 04 are where AgentRel is different.
Capability pillars
Twelve pillars. One adoption system.
Not a service menu. Each pillar covers one stage of the model and produces artifacts the next pillar depends on. Open any one to see what actually ships.
01Developer discoveryMake developers find you.
Technical buyers search differently, and increasingly they don't search at all — they ask. Discovery now has to work for indexes, answer engines, and generative systems at the same time.
Technical SEO
- Documentation SEO
- Developer search optimization
- Integration pages
- Technical landing pages
AEO — answer engine optimization
- Content structured so answer engines resolve product questions correctly
- Question-level coverage of real developer intents
GEO — generative engine optimization
- Visibility and authority inside generative AI systems
- Entity clarity and citation-worthy source material
LLM visibility measurement
- Product mentions
- Recommendations
- Comparisons
- Accuracy
- Citations
- AI-generated answers
Developer content
- Technical blogs
- Tutorials and guides
- Architecture articles
- Comparison content
- Research and engineering stories
02Documentation engineeringTurn documentation into a product.
Documentation is not a support artifact. It is the interface most of your users read before they ever touch the API — and now the interface machines read instead of the API.
Architecture
- Documentation architecture
- Getting started and quickstarts
- API references
- SDK documentation
- Versioning
Depth
- Tutorials and how-to guides
- Architecture guides
- Migration guides
- Troubleshooting and FAQ
- Security documentation
AI-native documentation
- Structured content and machine-readable references
- llms.txt where appropriate
- AI context packs
- Explicit constraints and complete code examples
- Structured schemas and agent workflows
03Developer experienceReduce time to first value.
The evaluation is won or lost in the first session. Every minute between landing and a working call is a minute a developer can spend on someone else's product.
Surfaces
- Developer portals
- API documentation
- SDKs and starter kits
- GitHub examples
- Postman collections
Hands-on
- API explorers
- Sandboxes and environment packs
- Interactive documentation
- Developer playgrounds
- CLI tooling
Reach
- Framework integrations
- Cloud integrations
Time to first value3 hours → 18 minutesIllustrative target, not a guaranteed result.
04Agent experienceYour next user may be an agent.
Traditional UX routes a human to a product. Developer experience routes a developer to an API. Agent experience routes an autonomous system through tools and protocols — and it fails in ways a UI review will never catch.
Design
- Agent experience audit
- Agent-ready API design
- Agent tool design
- Tool schemas and structured outputs
Control
- Agent instructions
- Agent workflows
- Agent permissions
- Human-in-the-loop workflows
Assurance
- Agent testing
- Agent observability
05MCP developmentTurn your APIs into tools AI agents can use.
If you already have a REST API, a GraphQL endpoint, an SDK, a SaaS platform, an internal service, or a database, the capability already exists. What's missing is a surface an agent can discover and call correctly.
MCP server development
- REST → MCP
- GraphQL → MCP
- SDK → MCP
- SaaS → MCP
- Internal systems → MCP
Tool design
- Tool naming and descriptions
- Input and output schemas
- Error handling
- Tool boundaries and context design
Resources
- Product information
- Documentation
- Data and files
- Knowledge
Security
- OAuth and API keys
- Scoped permissions and least privilege
- Audit logging
- Rate limiting
Testing
- Tool discovery
- Tool selection
- Parameter accuracy
- Execution success
- Error recovery
- Security
06Agent and protocol engineeringProtocols are the infrastructure. Business outcomes are the goal.
We evaluate which protocols actually earn their place in your architecture. Most products need two or three. Almost none need all of them.
MCP
- Model Context Protocol — AI-to-software tools and contextual capabilities
A2A
- Agent-to-agent communication and collaboration
A2UI
- Agent-driven and dynamic user experiences
x402
- Machine-to-machine and agentic payments
APIs
- REST, GraphQL and others remain foundational
Webhooks
- Event-driven agent workflows
SDKs
- Developer and AI coding-agent integration
Emerging
- Continuous evaluation of new standards and protocols
07Agentic commercePrepare your software for machine-to-machine commerce.
Some APIs and digital services can be consumed and paid for autonomously by software agents. Many cannot, and should not. We assess suitability before we build anything.
Readiness
- x402 readiness
- Payment-enabled APIs
- Usage-based APIs
- API monetization
Workflows
- Agentic payment workflows
- Machine-to-machine commerce
- Payment authorization
Controls
- Spending controls
- Transaction logging
08Agent security and governanceAgents execute real actions. Govern them like it.
The moment an agent can act, the risk surface stops being reputational and starts being operational. Every capability needs an explicit boundary and an explicit approver.
Permission model
- Agent permissions
- Tool authorization
- Least privilege
- Tool policies
- Agent identity
Human control
- Human approval gates
- Transaction limits
- Risk controls
- Sensitive action controls
Accountability
- Audit logging
- Authentication
- Authorization
Boundary example
- Auto: search, analyze, draft, recommend
- Approval required: pay, delete, transfer, publish
09Agent evaluation and observabilityDon't just make agents work. Measure whether they work.
An agent integration that succeeds in a demo and fails at 4% of real calls is a support problem you cannot see. Evaluation is how you find that before your customers do.
Evaluation
- Agent evaluation
- Tool evaluation
- Task success measurement
- Tool selection accuracy
Operations
- Agent tracing
- Cost and token monitoring
- Latency
- Failure analysis
- Human intervention rate
Sample trace
- Discover tool ✓ · Select tool ✓ · Validate input ✓
- Execute ✓ · Recover error ✓ · Task success 96%
10Developer educationTurn users into experts.
Education is the cheapest retention mechanism a software company has, and the only one that produces advocates as a byproduct.
Programs
- Developer academy
- Courses and learning paths
- Workshops and webinars
- Developer challenges and hackathons
Certification
- Certified developer
- Certified integration specialist
- Certified AI developer
- Certified architect
11Community and developer relationsBuild an ecosystem, not an audience.
A community manager who only posts announcements is a broadcast channel. A real one answers technical questions, finds product gaps, and turns recurring problems into content.
Channels
- Community strategy
- Discord and Slack
- GitHub Discussions
- Developer forums
- Community management
Programs
- Developer advocacy
- Ambassador and champions programs
- Meetups and conferences
- Hackathons, AMAs and office hours
What good looks like
- Answers technical questions
- Identifies product gaps
- Collects feedback and creates content from recurring problems
- Connects developers and grows advocates
12Technical content engineOne feature becomes an adoption system.
We don't create content for content's sake. We create technical assets that move developers toward successful adoption — and each one is derived from a real product capability.
Per feature
- Documentation, tutorial, blog
- Example, video, social
- Playground, GitHub, community thread
- MCP tool, academy module, FAQ entry
MCP and agent infrastructure
Don’t just expose an API. Expose a capability.
An API tells a machine what endpoints exist. A tool tells it what it can accomplish, what it is allowed to do, and what happens when it gets the arguments wrong. Those are not the same document.
If an agent can’t understand your product, it can’t use your product.
You already have this
Existing product surface
REST APIGraphQLSDKSaaS platformInternal serviceDatabase
AgentRel builds this
MCP layer
Tool schemasResourcesScoped authError contractsRate limitsAudit log
This is your new user
AI agent
DiscoversSelectsValidatesExecutesRecovers
Protocol matrix
Which interface does your software need?
Start from the business need, not the acronym. Select a need to see the interface it actually implies — and when that interface is the wrong answer.
Business need
Potential technology02 / 08
MCP
Model Context Protocol exposes your capabilities as tools an AI system can discover, select, and call — with schemas, boundaries, and scoped permissions.
Fits: Right when assistants and coding agents should operate your product directly.
Doesn’t: Premature if your API contract is still unstable or undocumented.
Not every product needs every protocol. AgentRel determines the right architecture based on your product and business model — and says so when the answer is “none of these yet.”
Developer experience transformation
From API to agent-ready platform.
The API was never the problem. Everything that was missing around it was. This is the shape of a full engagement.
Landing to first successful call
3 hours18 min
Illustrative concept, not a guaranteed result. Time to first value is the metric we baseline first, because it predicts almost everything downstream.
Before2 surfaces
- API
- Basic documentation
After11 surfaces
- API
- SDK
- Documentation
- Postman collection
- Playground
- Examples
- AI context
- MCP server
- Agent workflows
- Community
- Academy
Productized assessments
Agent Readiness Audit™
Ten dimensions, scored, with the gaps sequenced into a roadmap. It is where every engagement starts and the only honest way to find out whether protocol work is even the right next move.
What gets analyzed
- API readiness
- Documentation readiness
- AI readiness
- MCP readiness
- Agent tool readiness
- A2A readiness
- Payment readiness
- Security readiness
- Observability readiness
- Developer experience
Scored report and roadmap. No engagement required.
AgentRel readiness score72/100
Developer Growth Score™
How much friction sits between a developer finding you and shipping with you?
Get your score ↗
AI Visibility™
Do AI systems know your product, describe it correctly, and recommend it?
Measure AI visibility ↗
Agent Discoverability™
When an agent needs a capability you provide, can it find and successfully use you?
Test agent discovery ↗
Figures shown are illustrative sample output, not client results.
Technical content engine
One feature. Twelve adoption assets.
We don’t create content for content’s sake. We create technical assets that move developers toward successful adoption — each one derived from a real product capability, each one landing on a different surface.
One product feature
DocumentationTutorialBlogExampleVideoSocialPlaygroundGitHubCommunityMCP toolAcademyFAQ
More adoptionEducation and community
Turn users into experts. Build an ecosystem, not an audience.
Developer academy
Education is the cheapest retention mechanism a software company has, and the only one that produces advocates as a byproduct.
Programs
- Courses and learning paths
- Workshops and webinars
- Developer challenges
- Hackathons
Measured by
- Course completion
- Certifications issued
- Certified developers active
Certified DeveloperCertified Integration SpecialistCertified AI DeveloperCertified Architect
Community and developer relations
A community manager who only posts announcements is a broadcast channel. Ours answer technical questions, find product gaps, and turn recurring problems into content.
Channels
- Discord and Slack
- GitHub Discussions
- Developer forums
- Meetups and conferences
Programs
- Developer advocacy
- Ambassadors and champions
- AMAs and office hours
- Hackathons
Questions resolvedContributorsActive developers
Engagement model
Assess, architect, build, launch, operate, optimize.
Six phases, run in order. Most engagements stay in operate and optimize indefinitely, because adoption is a system to run, not a project to finish.
- 01
Assess
- DevRel and DX
- Documentation
- AI visibility
- Agent readiness
- Protocol readiness
- 02
Architect
- Developer roadmap
- Documentation architecture
- Agent architecture
- Protocol strategy
- Community strategy
- 03
Build
- Docs and SDKs
- Playgrounds
- MCP and agent tools
- Content
- Academy
- 04
Launch
- Developers
- Communities
- AI visibility
- Events
- Education
- 05
Operate
- DevRel and community
- Content
- Agent evaluation
- AI visibility
- Developer feedback
- 06
Optimize
- Adoption
- Time to first value
- Developer success
- Agent success
- Community growth
The engagement
DevRel as a service. A developer relations team you don’t have to hire.
Most software companies need documentation engineering, developer experience, agent infrastructure, and community — and none of those is a full-time hire on its own. We run the function as one team, at whatever depth the product actually needs.
- Take the whole function, or plug in beside an existing team
- Deliverables ship continuously — no 100-page report and a handshake
- Measured on adoption metrics you already track, not impressions
Starts with the audit. Run it yourself first if you’d rather.
Foundation
Startups establishing a first credible developer surface.
Scope per engagement
Growth
Growing software companies scaling adoption deliberately.
Scope per engagement
Agent Ready
Companies preparing their product for AI agents and protocols.
Most requested
Full DevRel
A complete external developer relations function.
Scope per engagement
Enterprise
Custom developer and agent infrastructure programs.
Scope per engagement
What we report
Metrics that describe adoption, not attention.
Developer
01- Time to first value
- Quickstart completion
- API activation
- SDK usage
- Integration success
AI
02- AI mentions
- AI accuracy
- AI recommendations
- AI integration success
Agent
03- Tool success
- Task success
- Agent discoverability
- Agent intervention rate
- Agent cost
Community
04- Active developers
- Contributors
- Community growth
- Questions resolved
- Ambassadors
Education
05- Course completion
- Certifications
- Certified developers
Impressions, reach, follower count, and MQLs are not on this list on purpose.
AgentRel Labs
We build with these protocols, not just about them.
Open source, Apache-2.0, no telemetry. The audit skills below are the assessment stage of our own service, given away — install them and score your surface without talking to us.
$ /plugin marketplace add agentrel/skills
agentrel/skills
Claude Code plugin. Five read-only skills that audit a developer surface and rank what to fix, plus a subagent for DX regressions in a diff.
shippedplugin- /agentrel:surface-audit
- /agentrel:ttfv-trace
- /agentrel:approval-gates
- /agentrel:docs-architect
- /agentrel:feature-to-assets
agentrel/toolkit
The same scoring as a CLI, so the audit runs on every merge instead of once. Exits non-zero below a threshold.
nextagentrel/mcp-kit
TypeScript primitives for building MCP servers with governance built in — approval gates, spend limits, audit trails.
plannedagentrel/x402-sandbox
Experiments in machine-to-machine settlement for metered APIs. Most products should not do this; we want to know which should.
experimentagentrel/a2a-examples
Worked examples of agents delegating across system boundaries.
experimentagentrel/a2ui-prototypes
Prototypes for agent-driven review and approval surfaces.
experimentgithub.com/agentrel ↗
One shipped, the rest published as they stabilize. Experiments stay labelled as experiments — we would rather archive one loudly than let it rot quietly.
Our vision
AgentRel OS
Where the services are heading: continuous software intelligence instead of periodic audits. One surface monitoring developer growth, AI visibility, agent readiness, documentation, community, and protocol health as they move.
Developer growth score
74
AI visibility
68%
Agent discoverability
62%
MCP health
99.2%
Documentation health
81
Developer funnel
3.4k
Agent task success
96%
Community intelligence
412
AgentRel OS is a product direction, not a shipping platform. Nothing above is available today — the modules describe what the services already measure by hand.
Why AgentRel
Four reasons this is not the same purchase.
Software-centric
We understand software products, APIs, SDKs, and the workflows developers actually have. The work starts inside the product, not around it.
Developer and agent
We optimize for both audiences at once, because the same artifacts serve both — and because optimizing for one usually breaks the other if nobody is watching.
Build, don't just promote
Documentation, SDK experiences, playgrounds, MCP servers, agent tools, communities, and academies. Deliverables, not decks.
Protocol-agnostic
We don't force a protocol. We identify the right technology for your business, and we say so when the honest answer is that you don't need one yet.
Traditional marketingAgentRel
CampaignsDeveloper systems
AdsDeveloper discovery
Generic SEOTechnical SEO + AEO + GEO
Blog postsTechnical adoption content
Landing pagesDeveloper experiences
Product videosTechnical demonstrations
Lead generationDeveloper activation
Marketing funnelAdoption funnel
Customer supportDeveloper community
WebsiteDeveloper portal
API documentationDeveloper + AI documentation
APIAPI + SDK + agent interfaces
ChatbotAgent tools
PaymentsAgentic commerce where applicable
We don’t market around the software. We build the infrastructure that makes the software adoptable.
Objections, answered
The questions CTOs actually ask first.
01Is AgentRel a marketing agency?
No. We build the artifacts developers and agents consume — documentation systems, SDKs, playgrounds, MCP servers, agent tools, communities, and academies. Campaigns are not the deliverable; adoption is.
02Does every product need MCP, A2A, or x402?
No, and we will tell you when it does not. We assess your product, business model, and ecosystem first, then recommend only the interfaces that produce a real outcome. Protocols are tools, not goals.
03Do you replace our developer relations team?
Either. Some clients outsource the whole function; others plug us in beside an existing team to cover documentation engineering, MCP development, or agent readiness specifically.
04How do you measure whether the work succeeded?
Against adoption metrics, not impressions: time to first value, quickstart completion, API activation, integration success, agent task success, and AI answer accuracy. We baseline every one of them during the assessment.
05What if our API is not ready for agents yet?
That is the usual starting point. The Agent Readiness Audit scores API, documentation, tooling, security, and observability, then produces a sequenced roadmap. Most engagements fix the API and documentation layer well before any agent work begins.
06Which stacks and platforms do you work with?
Any product with an API, SDK, or programmable surface — REST, GraphQL, SaaS platforms, internal services, and data systems. The interface layer we build on top is chosen per product.
07How long before we see anything shipped?
The assessment phase produces a scored readiness report and roadmap. Build waves ship incrementally after that — documentation and quickstarts first, because they move time-to-first-value fastest.
Start here
Is your software ready for the next generation of users?
Your product needs to be readable by people, runnable by developers, and callable by machines. The audit tells you which of those three you are missing.
Scored report and roadmap. No engagement required.
- 01Developers are becoming AI-assisted.
- 02Software is becoming agent-accessible.
- 03Payments are becoming machine-driven.
- →Your product needs to be ready.
AgentRel
Software is becoming increasingly discoverable through AI, built through AI, operated by AI, and consumed by agents. AgentRel exists to make software ready for that world.
Engineer adoption. Enable agents. Grow ecosystems.