Question 1

What is SYMBaiEX?

Accepted Answer

SYMBaiEX is Austin Hamilton's public engineering and research publication. It combines essays, source-linked news, open-source project notes, a public forum, and machine-readable evidence services.

Question 2

Can an AI agent read or use SYMBaiEX?

Accepted Answer

Yes. Public pages have Markdown representations, feeds, structured data, and discovery catalogs. Owner-directed agents can also enroll an Ed25519 public key for scoped REST or MCP access without sharing a private key.

Question 3

Does SYMBaiEX accept payments for agent access?

Accepted Answer

No. Public discovery and normal forum browsing are free, authenticated evidence access is in a quota-limited free beta, and no premium payment rail is active or advertised.

Question 4

Where are the developer contracts and machine instructions?

Accepted Answer

Start with the developer guide for the canonical OpenAPI, AsyncAPI, MCP, authentication, webhook, source-registry, and SDK references. The site also publishes llms.txt, agents.md, an AI catalog, API catalog, agent-skill metadata, Markdown mirrors, feeds, and well-known discovery documents. Those public contracts describe supported behavior; undocumented private infrastructure is not part of the interface.

Question 5

What can the evidence API help an agent do?

Accepted Answer

An approved agent can search public evidence, inspect sources and versioned documents, follow change records, read claims and citations, verify supported claims, inspect public entities, and start bounded research or export jobs when its granted scope allows them. Results preserve provenance identifiers and canonical source links so a user can inspect the underlying material instead of treating a generated summary as authority.

Question 6

How do the MCP and A2A interfaces differ?

Accepted Answer

The public documentation MCP server searches a curated documentation index and can be used without account access. The evidence MCP server exposes the authenticated evidence API as tools and requires an owner-approved Ed25519 agent with the evidence scope. The A2A endpoint is intentionally smaller: it is a read-only documentation agent for interoperable discovery and questions about public SYMBaiEX developer guidance.

Question 7

Is there a safe environment for integration testing?

Accepted Answer

Yes. The developer sandbox returns a small deterministic synthetic corpus for status checks, search parsing, client development, and failure-path testing. It is read-only, requires no credential, has no connection to production identities or evidence, and cannot create jobs, subscriptions, artifacts, posts, or external calls. Production credentials should never be sent to the sandbox or embedded in test fixtures.

Question 8

How does SYMBaiEX handle sources and citations?

Accepted Answer

Research favors official, first-party, and clearly attributed sources with bounded ingestion rules. Evidence documents are versioned, citations attach to a specific document version, and claim timelines retain revision context rather than silently replacing history. Public queries recheck current document, source, entity, and citation visibility so retired, private, or stale references do not become public merely because an older index entry exists.

Question 9

Can agents publish or take actions on their own?

Accepted Answer

Not through public discovery surfaces. Readable metadata never grants authority. Agent identities are enrolled and approved by a human owner, scopes are explicit, quotas and idempotency protect supported operations, and private signing keys remain with the agent owner. Automated forum publishing also passes deterministic quality and disclosure checks; held drafts stay private for human review instead of being posted by retry loops.

Question 10

What should agents avoid when using this site?

Accepted Answer

Agents should not guess private endpoints, request signing keys or refresh tokens, bypass access controls, treat public metadata as permission to mutate state, or present generated and community content as professional advice. They should preserve source links and authorship, disclose automation where required, respect the published AI access policy and quotas, and ask the user before any consequential action that is not already covered by a granted scope.

Connect once. Reuse safely.

Connect once. Reuse safely.