Manifest

Purpose

Action Network Worldwide (ANW) is a platform for practitioners and organizations working in locally led social and environmental development. It provides tools for discovering and sharing opportunities, coordinating training and learning, and matching people with the resources and partners they are looking for. The platform is operated by Weltweit – Gesellschaft zur Förderung lokaler Initiativen e.V., a registered non-profit association under German law (eingetragener Verein), founded in 2014 and registered at Amtsgericht Königstein im Taunus, VR 1327.

The people who use this platform are professionals, students, alumni, and organizations engaged in international cooperation – particularly those with ties to the Global South who work across geographic and institutional boundaries. The platform is not a social network in the consumer sense. It is a working infrastructure for people who need to find each other, exchange substantive information, and get things done together.

Mission & Vision

Our mission is to strengthen locally led social and environmental initiatives by supporting the people who can bridge contexts: international students and alumni from the Global South who study in Europe – particularly Germany – and remain connected to their home regions, communities, and institutions.

The cooperation systems we want to see are ones in which communities retain decision-making power over priorities, methods, and implementation, and in which transnational networks function as durable infrastructure – sustained pathways for knowledge, trust, and resources – rather than one-off project relationships that expire with the funding cycle.

Why Alumni and Diaspora Networks

Alumni and diaspora networks are central to our theory of change for three concrete reasons. First, they translate institutional opportunities – funding streams, standards frameworks, international partnerships – into locally workable steps, because they understand both sides of the gap. Second, they remain accountable to local stakeholders through long-term personal and professional relationships that outlast any single grant. Third, they reinforce continuity: when a project ends, the network stays. The platform is designed to support these networks, not to replace the relationships within them.

What We Build

The platform provides a set of tools for professionals and organizations in international cooperation:

• Profiles – individual and organizational, with structured fields for focus areas, skills, regions, and interests
• Opportunity discovery – a shared board for posting and finding funding calls, jobs, partnerships, training programs, and collaboration requests
• Matching – algorithmic suggestions based solely on profile data the user explicitly provides; no behavioral inference
• Networking – a member directory with configurable visibility settings and an interactive map for geographic exploration; currently over 1,000 members across 56 thematic spaces
• Organizational tools – team management, invite flows, and organizational profiles for associations and initiatives
• Collaborative workspaces – shared spaces for coordinating work, with optional Google Drive integration for document management; access is invitation-based and controlled by workspace coordinators
• Notifications – in-app and email alerts for relevant activity; every alert type can be individually enabled or disabled
• API access tokens – users can authorize external tools to read their data via token-based access; creation and revocation are fully under user control at any time

The platform codebase is open source. The matching algorithms are readable code – not opaque machine-learning models – and every formula used in any calculation is documented with cited sources. Infrastructure configuration is reproducible. There are no proprietary components that cannot be inspected or replaced.

How We Build

These are operating commitments, not aspirational values. We state them as what we do, because that is the only form that can be verified.

Your Data, Your Control

We collect the data needed to operate the platform and nothing beyond that. Profile information, opportunity listings, and workspace content are collected because users create them. Technical session data is collected because authentication requires it. We do not profile users based on behavior, infer attributes from usage patterns, or pass personal data to third-party analytics services. Self-hosted analytics tracks aggregate page views, not individual identity.

Matching uses only data you explicitly provide. If you do not fill in a field, it does not influence your matches. There is no inferred interest graph, no behavioral model, no scoring based on what you click or view.

Privacy settings give users control over what other members can see: profile visibility, whether an email address is shown, and location information can each be configured independently. You can delete your account and all associated data at any time. Deletion is processed, not deferred. We do not archive deleted accounts.

Notification preferences are fully configurable per alert type – no all-or-nothing switches. API access tokens give you a way to connect external tools to your account; you control creation and revocation at any time, and we do not share tokens with third parties.

Open by Default

The platform source code is publicly available. The matching algorithm is implemented in readable, documented code; there is no black-box model making decisions users cannot inspect. Every formula used in energy calculations, carbon estimates, or any other quantitative output is published with its variable values and source citations. No external dependency is integrated without an adapter interface that allows it to be swapped out. The platform does not depend on any single cloud provider, AI provider, database vendor, or authentication service in a way that would require rebuilding the system to change providers.

Human Decisions

Automated systems on this platform are tools for flagging and suggesting, not authorities that act on their own. Quality gates on opportunity listings flag content for review; they do not auto-reject. Publishing and moderation decisions are made by people. There is no gamification: no public scores, no leaderboards, no badges, no rankings. AI features assist with drafting, summarization, and matching suggestions – the user decides what to do with those suggestions.

No Lock-In

Infrastructure is designed to survive provider changes without rebuilding the platform. Authentication uses standard OIDC. The database is standard PostgreSQL. AI features use OpenAI-compatible APIs and can be pointed at any compatible provider. Every external dependency has an adapter layer: the provider can be replaced, and the platform continues to work. Data portability is a design requirement, not an afterthought. If you or your organization leaves the platform, your data leaves with you in a usable format.

Environmental Transparency

We track the energy consumption and CO₂ emissions of this platform because we think platforms should know what they cost to run. We do not claim that tracking reduces emissions – it makes them visible. What we do with that information is a separate and ongoing question.

Three Sources of Energy

The platform accounts for energy from three distinct sources:

• Infrastructure (Scope 2) – CPU utilization percentage, measured per minute via the DigitalOcean Monitoring API, multiplied by the server’s thermal design power (TDP) and the datacenter’s Power Usage Effectiveness (PUE) to produce watt-hours consumed by our compute
• Remote AI calls (Scope 3) – every call to a language model or embedding model is individually logged with exact input and output token counts; 16 instrumented call sites cover all AI features across workspace, opportunity, profile, and knowledge functions; energy is estimated from published watt-hours per million tokens for each model
• Network transfer (Scope 3) – exact bytes transferred, multiplied by the IEA estimate of 0.06 kWh per gigabyte

Scope 1 (on-site combustion) does not apply – we run on shared cloud infrastructure. User device energy is excluded from our accounting; we note the boundary rather than ignore it.

What Is Measured Exactly vs. Estimated

Three inputs are exact measurements taken at runtime: CPU utilization percentage, AI token counts, and network bytes transferred. These are not approximations. The remaining inputs are estimates with documented uncertainty ranges:

• Droplet TDP (thermal design power) – estimated ±30%, derived from Intel and AMD spec sheets and cloud performance benchmarks
• PUE (Power Usage Effectiveness) – estimated ±15%, using the DigitalOcean FRA1 facility industry standard of 1.2
• Grid carbon intensity – public data ±10%, from the Umweltbundesamt (UBA): 363 g CO₂/kWh for the German grid in 2024
• Remote AI Wh/token – research estimate ±50%, from the Stanford AI Index; this uncertainty will narrow as the EU AI Act requires providers to disclose per-inference energy data starting in 2027

We apply a 1.5× safety multiplier to the monthly CO₂ total before any offset calculation. We deliberately over-count rather than under-count.

Methodology: anw-carbon-v1

The methodology is named anw-carbon-v1 and is documented and version-controlled alongside the codebase. It follows GHG Protocol location-based accounting:

• Scope 2 – compute energy for our own infrastructure (purchased electricity)
• Scope 3 – remote AI API calls and network data transfer
• Scope 1 – not applicable (no on-site combustion)

Every controlled variable – TDP, PUE, grid carbon intensity, AI energy rates – is stored in an immutable versioned table. Variables are never updated or deleted; when a source changes, a new record is created with a new effective date and the updated source citation. Old records remain permanently. This means every past monthly calculation can be fully reconstructed: the exact formula, the exact variable values, and the methodology version are all linked to each monthly total.

Safety Margin and Offsets

A 1.5× multiplier is applied to the final monthly total before any offset figure is calculated. We would rather report a number that is too high than one that is too low.

A monthly CarbonLedger records, for each offset purchase: the offset provider, project identifier, kilograms of CO₂, cost in euros, proof URL, and third-party verifier. The offset infrastructure is built and active. Purchasing is currently manual – direct API integration with offset providers is planned but not yet implemented. Offset status is publicly visible per month as one of three states: pending, purchased, or verified. Offsets do not undo emissions; we will not present them as if they do.

Public Carbon Data

All carbon data is available without authentication via two public API endpoints:

• GET /v1/public/carbon – current annual totals, offset status, methodology version, and the full formula with all benchmark values in the response body; the response includes a formulas and benchmarks object so any caller can see exactly which numbers produced the figures shown
• GET /v1/public/carbon/history – monthly CO₂ history for charting and external use

Regulatory Alignment

We track two EU frameworks proactively – not because we are currently required to, but because we expect to be and because the discipline they require is independently useful:

• EU AI Act Art. 40 – per-call energy measurement and methodology versioning are already live ahead of the 2027 requirements for AI system providers
• EU EED Article 12 – datacenter metrics (PUE, renewable energy percentage, grid carbon intensity) are snapshotted monthly in the carbon ledger; our shared infrastructure falls below the mandatory 500 kW reporting threshold, but we report anyway

Compliance & Standards

We follow these frameworks as design choices. We hold no certifications from any of them, and we do not claim the authority of any standards body.

• GDPR – data protection by design and by default, applied throughout the platform architecture. The platform is operated by a German legal entity and data is processed in accordance with EU law. We treat GDPR as a floor, not a ceiling.
• EU AI Act – per-call energy instrumentation and immutable methodology versioning are already implemented ahead of the 2027 requirements. Matching features use only user-provided profile data; there is no behavioral inference. Match scores are internal signals, not published rankings.
• GHG Protocol (Scope 2 & 3, location-based) – the methodological basis for our energy and carbon accounting. We use the location-based method because market-based accounting for shared cloud infrastructure involves assumptions we cannot independently verify.
• EU EED Article 12 – datacenter energy reporting metrics are tracked and snapshotted monthly even though our infrastructure is below the mandatory reporting threshold.

These are not badges. They are the frameworks against which we audit our own practices.

What We Don’t Do

These are hard limits, not aspirations. We state them plainly because clarity about what a platform will not do is as important as what it will.

• We don’t sell data. No user data – individual or aggregate – is sold to any third party for any purpose.
• We don’t run advertising. There is no advertising on this platform, no paid promotion in feeds, and no sponsored content. The platform is funded by membership fees and grants, not by attention economics.
• We don’t profile users based on behavior. Matching uses only the profile information users explicitly enter. We do not infer interests, attributes, or intent from usage patterns, click behavior, or viewing history.
• We don’t claim our carbon tracking reduces emissions. Measurement and reduction are different things. We measure because we think it is the responsible minimum. We do not present tracking as an environmental achievement.
• We don’t certify organizations or endorse opportunities. Listings on this platform are posted by members. We do not verify claims made in opportunity listings, and appearing on the platform does not constitute endorsement by ANW or by Weltweit e.V.
• We don’t operate as an employer or recruiter. We are not a party to any employment or contracting relationship that originates on this platform. We provide the infrastructure; the relationships are between the people using it.
• We don’t lock users into our platform. Data portability is a design requirement. Users can export their data. Every external integration is built with a replaceable adapter. There is no deliberate friction preventing users from leaving or moving their data elsewhere.

History

Action Network Worldwide has been built incrementally since 2014 by a small international team. The milestones below describe what actually happened, in order.

• 2014 – Weltweit e.V. founded in Germany by an international team with professional experience in international cooperation
• 2020 – Launch of longer-format online academies and applied training in project planning and monitoring
• 2021 – Expansion of thematic workshops covering monitoring and evaluation, storytelling, partnership management, and sustainable finance
• 2023 – Launch of the Action Network Organizations Alliance, a structured network tier for institutional members
• 2024 – Network reaches 895 members and 56 thematic spaces
• 2025 – Network grows to 1,016 members
• 2026 – Platform rebuilt with member profiles and organizational pages, matching engine, collaborative workspaces, Google Drive integration, in-app and email notifications, API access tokens, and an interactive member map

Contact

Questions about this manifest, the platform, or our operating principles can be submitted via the support page. We read everything and respond to substantive questions.

The technical manifest – covering system architecture, energy accounting formulas with full variable tables, database schemas, and API specifications – is published alongside the open-source codebase. If you are reviewing our methodology or auditing our claims, that is the right starting point.

Weltweit – Gesellschaft zur Förderung lokaler Initiativen e.V.
Amtsgericht Königstein im Taunus, VR 1327
actionnetwork.world