Author your BCP. Three ways in.

A guidelines doc is written for humans. BCP is written for agents. It is hierarchical, versioned, queryable, and readable by every LLM, vendor platform, and consumer agent in your stack.

Your doc tells one designer how to pick a color. Your BCP tells a thousand agents at once what you do and don't stand for. That includes ChatGPT describing you, your media-buying agent making a call, and your creative-gen tool writing copy. Different artifact for a different job.

You can, and you should expect it to work poorly. PDFs are unstructured, session-bound, and re-uploaded per chat. Every agent on every platform has to re-parse it, and every conversation starts from zero.

BCP is structured markdown at a public URL with a canonical path. Once it's there, every agent can fetch it, cache it, diff it when you update, and act on it without a human in the loop. It's the difference between emailing a style guide and publishing an API.

ChatGPT, Claude, Perplexity, and Gemini are already describing your brand right now. They use whatever signals are cheapest to reconstruct: a scraped product page, a five-year-old press release, a Reddit comment, or a competitor's positioning. The gaps get filled by hallucination.

Every day without a BCP is a day of someone else authoring your representation. The cost of doing nothing isn't zero. It is a slow drift of your brand into whatever the models happen to guess.

You. Every tier produces a Git repo you own, on your GitHub (or any Git host), served from your domain. Tacit doesn't custody your brand data. The spec is CC BY 4.0, the code is MIT. There is no lock-in by design.

Three tiers, all available today. Free: fork our public template and publish your own BCP with no Tacit involvement. $1,500 Expert Session: 90 minutes with a senior strategist who helps you draft your root brand.md and one priority daughter file. $15,000 Encoding Sprint: one week, full seven-file tree, authored and shipped to your domain.

For larger multi-brand or enterprise scope, we quote custom engagements. Forge (the self-serve wizard) is in development and will be free when it ships.

Honest answer: BCP is v0.1, and native integrations with vendor platforms are being built through 2026. Today, BCP works as a system-prompt paste into any LLM, and its boundaries file maps directly to IAB 3.0 and GARM Brand Safety Framework fields that brand-safety platforms already consume.

The long-term bet: vendor platforms adopt BCP because reading one structured file beats reconstructing brand intent from account-rep notes and onboarding interviews. Early vendor conversations are in progress.

BCP is versioned infrastructure, not a static deliverable. Every edit is a Git commit. Every file has a last_updated date and a revision hash. Your agents re-fetch and re-cache on the cadence you set.

The problem isn't change. It's drift. The monitoring layer we're building will surface the gap between what your BCP declares and how agents actually describe you. That's the part that turns a file into a living system.

Only if you go fully self-serve and push the template to your domain yourself (a dev task). The Expert Session and Encoding Sprint are both fully done-for-you. We author the tree, validate it, and ship it to your domain. Non-technical teams can hand us the login and receive a published BCP.

When Forge ships, even the self-serve path will be a wizard, not a repo fork.

For the same reason platforms adopted OAuth instead of proprietary login, and Schema.org instead of platform-specific markup: open standards free vendors from doing the onboarding discovery for every new brand, and they signal interoperability to customers.

Every brand wants to author its identity once and have every platform consume it. Every platform wants onboarding friction to drop. Open standards are how those two goals meet.

At the simplest level: fetch {brand}.{tld}/.well-known/brand.md, parse the markdown and YAML frontmatter, and map BCP fields to your internal schema. Reference integration patterns and validators are shipping as SDK samples.

If you're a vendor platform exploring integration, reach out directly. We will scope it as a joint design conversation.

BCP v0.1 is new. Tacit itself is the reference implementation. Our brand is published at hellotacit.ai/.well-known/brand.md. Early partner brands are being encoded now. We will name adopters publicly once partnerships are live and logos are approved. We don't claim partnerships we haven't signed.

BCP's boundaries file maps directly to IAB Content Taxonomy 3.0 and GARM Brand Safety Floor and Suitability Framework. Tacit-specific extensions (like positive adjacency and voice-level constraints) are namespaced under extensions: so vendor platforms can consume the standard fields cleanly and optionally read the extensions.

Three things. The reference implementation is live and dogfooded (the company selling the spec publishes its own BCP at its own domain). The licensing is real | MIT code, CC BY 4.0 spec | and the code is public on GitHub. And the pattern is proven: AGENTS.md and MCP used the same well-known-path convention and are already seeing real adoption.

Standards die when the maintainer doesn't use them, when the license is vague, or when there's no integration path. BCP is trying to fail at none of those.

They share the same structural pattern: a well-known path, hierarchical markdown or JSON, and developer-native tooling. Different domains. AGENTS.md is codebase-level instructions for coding agents. MCP is a protocol for tool and resource use. Schema.org is structured data for search engines.

BCP is the brand-identity layer. Together, they form the agent-readable web: what to build, what to do, what you look like.

Yes. Entirely. Fork the template on GitHub, edit for your brand, publish at your domain. The spec is CC BY 4.0, the reference code is MIT. We don't gate access to the standard.

Tacit makes money from authoring tools like Forge, consulting via Expert Sessions and Encoding Sprints, and eventually monitoring. We do not gatekeep the spec.