Atomic claims
Each Signal is a claim with evidence, confidence, entities, dates, source metadata, and domain classification.
Agents do not need another headline feed. They need source-grounded change in a format they can discover, inspect, pull, diff, save, and route. Synorb Core MCP exposes the Temporal Context Graph through Source Channel inventory, live Stream workflow tools, high-throughput relevant Manifest pulls, and reusable Beacon tools for authenticated users.
Core MCP: profile + 7 workflow tools · Monthly quota + per-call caps · Relevance controls
A news result tells an agent that something exists. A Manifest gives it the source event, the extracted claims, the compact Brief, stable IDs, and provenance in one object. Core MCP serves Signal + Brief plus source metadata. Full source payloads are Enterprise-only and are retrieved via REST.
Each Signal is a claim with evidence, confidence, entities, dates, source metadata, and domain classification.
Each Brief keeps the important claims together so agents and humans can review the source event without parsing raw prose again.
Each Record keeps joins, lineage, replay, and warehouse-grade use possible when the same event enters downstream systems.
Recent Core MCP upgrades that affect every synorb-manifests response.
Every Manifest carries a compact citation entry — citation_id, headline, stream, source channel, published date, and canonical URL. A top-level deduped list and ResourceLinks accompany the structured payload so agents can cite without parsing bodies.
Each Manifest reply starts with a deterministic quota_header. synorb-profile, synorb-stream-search, synorb-catalog, synorb-details, synorb-configure-beacon, synorb-save-beacon, synorb-beacons, and synorb-archive-beacon are zero-quota; synorb-manifests bills only returned on-topic Manifests. Normal MCP usage is governed by monthly Manifest quota, per-call page/target caps, date-window/access gates, and relevance controls, not tiny per-minute caps. Emergency buckets are high by design: discovery tools sustain 10 requests/second with 300-request bursts, Manifest/content tools sustain 5 requests/second with 240-request bursts, and full Records sustain 2 requests/second with 120-request bursts.
Explicit source_channel_names, exclude_tag_names, exclude_tag_ids, exclude_query_terms, exclude_source_channel_ids, and exclude_home_domains on synorb-manifests. Profile and manifest payloads expose index_freshness and degraded_mode so agents can tell users when answers reflect a stale window.
Core tools advertise output schemas. Manifest pulls can stream progress to clients that support MCP progress, and tools provide completions for common filters such as domains, media formats, Streams, source channels, and tags.
Connect once to Core MCP, let the agent check quota and date-window context with synorb-profile when needed, handle stream/source/podcast availability questions with synorb-stream-search, discover candidate Streams by topic with synorb-catalog, inspect filters and coverage with synorb-details, pull Manifests with synorb-manifests, and manage reusable monitoring queries with synorb-configure-beacon, synorb-save-beacon, synorb-beacons, and synorb-archive-beacon. Before saving, the agent should improve obvious metadata: a short useful name, plain-English description, and 1-5 lowercase tags, while keeping stream/source/topic/entity scope faithful to the proposal. Beacons are available on every plan with per-plan caps (Starter 10, Individual 50, Professional 250, Startup 1,000, Enterprise unlimited); if the account is at its cap, list with synorb-beacons, ask which Beacon to archive, and use synorb-archive-beacon. Exact source surfaces can go direct with source_channel_names, for example ['8k'] for SEC Form 8-K, ['earnings-call-transcripts'] for transcript-only workflows, or policy/court slugs such as ['federal-court-opinions']. If requested source names do not resolve, MCP returns zero Manifests without consuming quota instead of broadening to adjacent sources. Advanced MCP is a separate URL for configured workflows that need lower-level access.
Answer "what streams, sources, or podcasts do I have access to?" with ranked matches, typo correction, and scope-labeled counts. Read-only; does not consume Manifest quota.
Search by topic, entity, source type, or domain and get the Streams most likely to match the user's intent.
Read coverage, source channels (with publisher name, category, content types, URL, and agent-optimized description), home domains, tag types, date ranges, media formats, and safe query options. Source channel enrichment means agents understand what a stream covers without a separate lookup.
Apply filters, inspect counts, retry with broader Streams when needed, and return Signal + Brief plus source metadata. Full Records are Enterprise-only via REST.
Turn a tracking/query intent into a reusable configuration with streams, source channels, topics, and prompt guidance.
Create the account-level Beacon with refined obvious name, description, and tags, then return beacon_id when the user asks to save it.
List saved Beacons so agents can run the right reusable configuration on demand.
Archive saved Beacons when the user asks for cleanup, when an existing Beacon is stale, when a plan cap is reached and the user chooses one to archive, or for approved eval cleanup. Archive calls do not consume Manifest quota.
Agents need to know where a claim came from before they can act on it. Synorb keeps source name, URL, published date, captured date, Stream routing, and lineage together so the agent can trace a Manifest back to the watched surface.
{
"manifest_id": "1777525429698648000",
"stream_names": ["mayfield"],
"source": {
"name": "Mayfield Blog",
"media_format": "text",
"published_date": "2026-04-30"
},
"lineage": {
"source_channel_id": "mayfield-blog",
"record_id": "1777525429649909800",
"brief_id": "1777527738510498300"
},
"delivery": {
"interfaces": ["REST", "MCP", "webhook", "S3"]
}
}
A single agent may need all three. Signals drive reasoning and alerts. Briefs fit context windows and operator review. Records keep the original source event structured enough for joins, replay, and audit.
Use Briefs when an agent needs compact narrative without losing the claims it was built from.
Use Signals when the workflow needs claim-level matching, routing, confidence, and evidence.
Use Records when the data needs stable IDs, source metadata, warehouse joins, and replay.
MCP is the right surface when an agent needs to reason through intent, discover Streams, inspect filters, run Manifest queries, observe the result, and refine. For interactive agent workflows, MCP users and hosts should keep a persistent Streamable HTTP MCP session/connection warm across tool calls. Authenticated MCP supports high-throughput relevant Manifest pulls. REST/Fast API is the explicit path for stateless per-call usage, scheduled jobs, tests, dashboards, backend-owned product paths, deterministic polling, and REST-shaped contracts.
Give MCP-compatible clients and hosts the Core Streamable HTTP URL and MCP token. Use Authorization: Bearer for new integrations. Initialize once for the user or workspace, keep the session warm across tools/list and follow-on tool calls, and reuse it for the agent loop. Cold sessions and one-shot initialize/call/close flows are supported for compatibility and diagnostics, but they are not the normal low-latency SLO path. Usage is primarily bounded by monthly Manifest quota, per-call page/target caps, date-window/access gates, and relevance/off-topic suppression.
Use REST/Fast API for stateless per-call usage, deterministic polling, tests, dashboards, data jobs, backend-owned product paths, and integrations that need REST-shaped request/response contracts.
The Starter tier includes 1,000 manifests per month on monthly delivery. Streams scale by manifest volume and refresh cadence. Firehose is for platforms that need everything Synorb writes, the moment it is written.
Synorb is a temporal context graph for AI systems. News and source events are part of what enters the graph, but the product unit is the Manifest.
No. Core MCP is the agent-native discovery interface for hosts that can keep a warm interactive session. Advanced MCP, REST/Fast API, webhooks, and S3 are available for configured, stateless, scheduled, or application-owned delivery patterns.
Credentials return an API key, secret, MCP token, Streamable HTTP Core MCP config, tokenized connector URLs for compatible hosts, and Advanced MCP details. From there the agent should ask synorb-stream-search when the user is exploring inventory or a specific source, run synorb-catalog to find Streams by topic, inspect candidates with synorb-details, then retrieve with synorb-manifests. When the user asks to save/reuse a tracking query, call synorb-configure-beacon, improve obvious name, description, and tags while keeping scope faithful, and then call synorb-save-beacon to return beacon_id. Use synorb-beacons to find saved Beacons. Beacons are available on every plan with per-plan caps (Starter 10, Individual 50, Professional 250, Startup 1,000, Enterprise unlimited); if the account is at its cap, ask which Beacon to archive and use synorb-archive-beacon. If the user names an exact source surface such as SEC Form 8-K, earnings-call transcripts, or a known policy/court source slug, the agent can call synorb-manifests directly with source_channel_names and should not substitute adjacent source surfaces.
Starter tier: 1,000 manifests per calendar month on monthly delivery. Agents can self-provision credentials and MCP config with one request. Humans can use the credentials page and receive the same key, secret, MCP token, Streamable HTTP Core MCP config, tokenized connector URLs for compatible hosts, Advanced MCP details, and schema PDF by email.
curl -s https://synorb.com/connect
Returns: api_key, api_secret, mcp_token, Core MCP config, tokenized connector URLs, Advanced MCP details.