Pheromind
Pheromind is a free multi-agent development framework that orchestrates 30+ autonomous AI agents through a Neo4j knowledge graph and the SPARC methodology, carrying a software project from specification to verified delivery on your own infrastructure.
What is Pheromind?
Pheromind is a free framework for autonomous software development built on what its author calls AI swarm orchestration. It was created by Chris Royse and is published through an official GitHub repository that carried 379 stars, 127 forks and 113 commits when we looked. Its central idea is borrowed from ant colonies: the agents do not talk to each other directly, they coordinate indirectly through digital pheromone trails, a mechanism known as stigmergy.
The work is organised as a hierarchy. An Uber-Orchestrator sets strategy, phase orchestrators own each stage, and 30+ specialised worker agents do the production, among them a Research Planner, a Knowledge Weaver, a Goal Clarification Orchestrator, a Devil's Advocate Critical Evaluator, a Test-Driven Developer, a BMO Multi-Oracle Verifier, a Chaos Engineer and a Failure Pattern Learner. The site explains the structure with a construction-company analogy: a CEO, site managers, trades, a plan room, a safety inspector and a client.
Memory is the second pillar. Everything the swarm decides lives in a Neo4j knowledge graph presented as the single source of truth: decisions, failure patterns, architecture choices and end-to-end traceability. The graph is split in two. A "projectmemory" graph holds the why and the what and indexes the artefacts, while a "model" graph holds the how, mapping every function, class and variable with CALLS, USES and IMPORTS relations.
Delivery follows the SPARC cycle, for Specification, Pseudocode, Architecture, Refinement and Completion, and every phase is test-first. Verification is handled by BMO, for Behavior-Model-Oracle, a suite of functional, visual, performance, consistency and metamorphic oracles that returns a confidence score between 0.0 and 1.0.
What comes out is working code with its tests and its documentation, each traceable back to the specification that produced it. The positioning is openly David versus Goliath: give a solo developer or a two-to-five-person team the output of a far larger engineering organisation. Maturity is the caveat. The project is young, distributed through a private community rather than versioned public releases, and the publisher's own pages disagree on its scale, with 30+ agents in the body of the site against 50+ in the meta description.
What it does
- Orchestrates 30+ specialised AI agents in a hierarchy, from an Uber-Orchestrator down to phase orchestrators and worker agents
- Maintains a Neo4j knowledge graph as the single persistent memory of a project
- Runs the full SPARC cycle, from specification through pseudocode, architecture and refinement to completion
- Writes acceptance tests before the code and verifies its own output through the BMO oracle suite
- Analyses and refactors legacy code down to function, class and variable level
- Generates technical documentation as the project advances
- Injects faults to test resilience and stores failure patterns back into the graph for later projects
When to use Pheromind / When not to
A quick filter to help you decide if Pheromind is the right fit.
When to use Pheromind
- Solo developers and small teams of two to five people who need to ship working software without hiring more engineers
- Startup founders building an MVP, the use case the publisher pushes hardest, with a three-day MVP claimed on the site
- Engineering teams modernising legacy codebases, covered by a dedicated code-refactoring solution page
- QA teams that want acceptance tests written before the code and generated automatically by the swarm
- Developers comfortable with Docker, Neo4j and a local install, willing to steer agents rather than write every line themselves
When not to use Pheromind
- Non-technical users looking for a ready-made web app: nothing runs in a browser, the framework has to be installed and configured
- Teams that need contractual guarantees: there is no DPA, no subprocessor list and no published SLA outside the quote-based enterprise offer
- Organisations under strict compliance audit: "SOC 2 Type II" and "GDPR Ready" are marketing badges, with no trust or security page and no report behind them
- Anyone who needs an API: no API documentation and no /docs route exist anywhere on the site
- Buyers who require an identified vendor: no legal entity, no postal address, and a single personal email address covering legal, GDPR and support
How to use Pheromind
A typical end-to-end flow, from setup to results.
- Join the Insiders community at insiders.pheromind.ai, or the Discord server alongside it, since the framework files are shared in the group announcements and there is no public download and no SaaS account to open
- Download the Pheromind files and install the framework locally on your own machine
- Install and configure Neo4j, which holds the project knowledge graph, and have Docker available in your stack
- Optionally book the paid Pheromind Guided Setup, a one-hour session held every Monday at 7 PM EST for a one-off 49 USD, if you would rather be walked through the installation
- Open a project with the Goal Clarification step, where you state the vision and the orchestrator interrogates it before any code is written
- Let the swarm run the five SPARC phases, from specification and pseudocode through architecture and refinement to completion
- Review and approve the milestones the orchestrators hand back: you supply the direction and the validation, the agents do the production work
- Query the knowledge graph to trace any line of code back to the requirement that produced it, and to read the failure patterns the swarm has recorded
- Allow two to three days to get the basics and two to four weeks for full command with the structured training, according to the publisher's FAQ
Pros & Cons
Pros
- Genuinely free, with commercial use allowed and a single obligation: crediting Pheromind
- The complete source code is accessible, deployable and modifiable under the community Code Credit License
- Self-hosted by design, so the code the agents work on stays on your own infrastructure with no SaaS intermediary
- Persistent project memory in a knowledge graph addresses a real problem, the loss of context between sessions
- Quality is built into the process rather than bolted on: TDD, adversarial critique, chaos engineering and a security review module
- Language- and stack-agnostic, thanks to pseudocode blueprints rather than fixed templates
- The architecture is documented in depth, through a technical blog and two ResearchGate whitepapers, and the community is active on Discord and GitHub Discussions, with 379 stars on the official repository
Cons
- Installation and configuration are not trivial: Neo4j and Docker to set up, files to fetch from a private community, and no graphical interface at all
- No API documentation and no /docs route anywhere on the site
- Marketing figures are unverifiable and sometimes contradict each other, with 30+ agents claimed in the body of the site against 50+ in the meta description
- Case studies and testimonials, attributed to Sarah Chen at PayFlow, Mike Rodriguez at ShopSmart and Dr Jennifer Park at MedFlow, cannot be checked, and some claims are extreme, such as a 50,000,000 USD government contract or 1M+ users in 90 days
- "SOC 2 Type II Compliant" and "GDPR Ready" are displayed as badges with no trust or security page, no report and no named auditor, and "MIT Licensed Core Components" is claimed while no LICENSE file appears in the official GitHub repository listing
- No identified legal entity, no postal address and one personal email address behind legal, GDPR and support; the privacy policy and terms of 20 June 2025 are generic and cover the website, not the product
- A young project distributed through a private community rather than public versioned releases: domain registered on 15 May 2025, first web archive capture on 6 September 2025, and seed funding still being sought
Pricing & Plans
Pheromind itself is free. The pricing page shows "Your Price: $0" for the framework and the FAQ confirms that it is completely free with no strings attached, the publisher stating that its revenue comes from training rather than licences. There is therefore no entry price for the software and no paid tier of the tool. Everything the pricing page charges for is a service, not a plan: a one-off 49 USD Pheromind Guided Setup session, 99 USD per month for the Weekly Live Q&A with Chris, 300 USD per hour for one-to-one consultation, 15,000 USD for a one-week AI Development Proof of Concept, 25,000 USD for a two-week Developer Enhancement Program, and a quote-based Enterprise AI Integration. Promotional banners elsewhere on the site advertise a 49 USD Pheromind 101 course, a 997 USD MVP workshop against a struck-through 2,497 USD, and a community membership announced at 39 to 79 USD per month. Money-back guarantees are displayed on the Proof of Concept and on the Developer Enhancement Program.
- the framework itself
- with full access to the Pheromind codebase
- community support and commercial projects allowed
- the only obligation being to credit Pheromind. Every paid entry below buys training or consulting services
- not a tier of the software
- a one-hour session covering complete installation and configuration
- starting advice and access to the Monday sessions
- weekly one-hour live sessions
- shared work on projects
- direct access to Chris and community interaction
- personalised guidance
- technical troubleshooting and strategic advice
- limited platform access
- one delivered project
- 16 hours of basic SPARC training
- a benchmark report
- architecture recommendations and a productivity analysis
- full access to the 30+ agents
- Neo4j installation
- 40 hours of SPARC training
- custom agent configuration
- integration with existing tooling and weekly reviews
- private cloud deployment
- custom agents
- enterprise Neo4j
- team lead certification of 50+ hours
- developer training of 100+ hours
- a dedicated success manager
- dashboards
- a custom SLA and 24/7 support
Data, GDPR & hosting
A consolidated view of how Pheromind handles your data.
GDPR overview
The word "GDPR" appears nowhere in the privacy policy or the terms. Its only occurrence on the entire site is a homepage marketing badge reading "Enterprise Security Certified - SOC 2 Type II Compliant - GDPR Ready". That is a claim, not a certification: there is no trust or security page, no audit report and no named auditor. The privacy policy does carry a "Your Rights" section covering access, correction, deletion, opting out of marketing communications and portability. Everything else a GDPR file needs is missing: no legal basis, no retention period, no data protection officer, no Article 27 EU representative, no data processing agreement, no subprocessor list and no international transfer clause. The only channel to exercise those rights is chris@pheromind.ai. The terms name no governing law, referring only to the jurisdiction in which Pheromind operates.
Who owns the data?
The privacy policy, dated 20 June 2025, covers the website and its associated services only, meaning the newsletter, contact forms and courses. It claims the name, email address and contact details you submit, plus automatically collected IP address, browser type, pages visited and time spent. It says nothing about the code the agents process. The terms state that all content, features and functionality of Pheromind's services are the exclusive property of Pheromind. On the product side the community tier grants full access to the codebase, to deploy, modify and scale as you need, under a "Code Credit License" whose only obligation is to credit Pheromind, while the Proof of Concept offer promises complete project source code ownership.
Reuse rights
The declared purposes are providing and improving the services, communicating about products and services, sending newsletters with consent, answering enquiries and support requests, analysing site usage and meeting legal obligations. Data is shared with providers that help operate the site, with authorities on legal request, and with business partners only on explicit consent. The policy states plainly that personal information is neither sold, traded nor otherwise transferred to third parties without consent. Cookies and tracking technologies are used for experience, traffic and referral analysis, with browser settings offered as the control; a consent banner exposes four categories, Necessary, Analytics, Marketing and Functional, and Google Analytics and Hey Oliver tags are present in the page, alongside a Supabase backend visible in the site bundle. Nothing in the privacy policy, the terms or the FAQ ever mentions training models on customer data, in either direction. On the reuse side, the community licence lets you run Pheromind on commercial projects without asking permission, the single condition being to credit Pheromind.
Data retention & training
Hosting summary
No hosting information is published. There is no trust or security page, no data centre location, no region and no named infrastructure provider anywhere on the site, in the privacy policy or in the terms. The site itself sits behind Cloudflare's anycast network: the resolved IP address 185.158.133.1 corresponds to a Cloudflare node observed in Poland, which says nothing about where any data is actually stored, since an anycast address is not a hosting jurisdiction. A Supabase backend, handling authentication, storage and edge functions, is visible in the site's JavaScript bundle but is never documented publicly. AWS and Azure appear on the site only as product integrations, never as hosting locations, and should not be read as such. One structural point works in the user's favour: the framework is self-hosted. It is downloaded and deployed on the user's own machines, with Neo4j and Docker, so the code the agents process stays on the user's infrastructure rather than on the publisher's. What the publisher's own website collects, meaning form submissions, newsletter sign-ups and analytics, remains hosted somewhere undisclosed.
Things to keep in mind
Risks and trade-offs to weigh before adopting Pheromind.
- Compliance badges are not evidence. "SOC 2 Type II Compliant" and "GDPR Ready" sit on the homepage with no trust or security page, no report and no named auditor, and the word GDPR appears in neither the privacy policy nor the terms. Do not read them as certifications
- "MIT Licensed Core Components" is claimed on the site, yet no LICENSE file appears in the official GitHub repository listing. Check the licence terms yourself before shipping commercial work built with it
- The publisher contradicts itself on scale, with 30+ agents in the body of the site against 50+ in the meta description and one case study
- Case studies and testimonials report extreme outcomes, including a 50,000,000 USD government contract and 1M+ users in 90 days, and cannot be verified; the on-site ROI calculator produces self-declared savings, such as a 97% reduction in delivery time, with no audited method
- There is no legal entity, no postal address and a single personal email address handling legal, GDPR and support, while the terms name no governing law. Recourse in case of a dispute would be difficult to organise
- The privacy policy and the terms cover the website only. Nothing anywhere describes what happens to the code the agents process, so proprietary or regulated codebases deserve caution until you have verified your own deployment yourself
- Delegating a whole specification-to-delivery cycle to a swarm concentrates human effort on approving milestones. Without deliberate reading of the output, a team can end up shipping and maintaining code nobody on staff actually understands
Setup & Integrations
Technical difficulty
High for a non-developer, moderate for an engineer. There is no sign-up and no web interface: you join the Insiders community, download the framework files from the group announcements and deploy them yourself, with Neo4j to install and configure and Docker in the stack. The publisher sells a one-hour Guided Setup session at 49 USD, which in itself suggests that getting started is not immediate, while its FAQ claims no prior experience is required and puts the basics at two to three days. In practice the real audience is developers who are comfortable on the command line.
Integrations
Behind Pheromind
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Pheromind.
Frequently asked questions
Is Pheromind really free?
How do I get access to the framework?
How is it different from GitHub Copilot or Cursor?
Which programming languages does it handle?
How is the quality of the generated code controlled?
Is there an API or a mobile app?
How long does it take to become productive?
What support do free users get?
Is my code used to train models?
Is a data processing agreement or a subprocessor list available?
Should you pick Pheromind?
Pheromind proposes something genuinely original and, unusually for this segment, documents it properly: stigmergy borrowed from ant colonies, a Neo4j knowledge graph as persistent project memory, the SPARC cycle and BMO verification with confidence scores. The technical writing behind it, a blog and two ResearchGate whitepapers, goes well beyond the usual landing page.
It is also genuinely free. Not a trial, not a freemium ladder: the framework is given away under a community licence whose only condition is to credit the publisher, and it is self-hosted, so the code your agents work on never leaves your infrastructure. Everything with a price tag, from the 49 USD setup session to the 25,000 USD delivery programme, buys human services rather than access to the software.
The counterweight is governance. The domain was registered in May 2025, the first web archive capture is September 2025, and the roadmap still lists securing seed funding as a current objective. No legal entity is published, no postal address exists, and a single personal email address answers legal, GDPR and support questions. The SOC 2 Type II and GDPR Ready badges have no report, no auditor and no trust page behind them. The marketing is loud, with case studies reporting outcomes nobody can check, an ROI calculator producing self-declared savings, and the publisher's own pages disagreeing on how many agents there are.
Read the two halves separately. For an autonomous technical team that can install Neo4j and Docker and judge the output for itself, Pheromind costs nothing to try and the architecture is worth the reading. For a regulated enterprise deployment, everything around the code, from compliance claims to the identity of the counterparty, needs to be instructed carefully before anything is signed.
- Choosing a selection results in a full page refresh.
- Opens in a new window.