{
"markdown": "# SAIHM MCP Server\n\n**Portable memory for AI agents.** Your agent remembers what matters — across\nsessions, across models, and across vendors — and you can share it, revoke it,\nor erase it for good.\n\n[](https://www.npmjs.com/package/@saihm/mcp-server)\n[](https://www.bestpractices.dev/projects/12898)\n· Apache-2.0\n\nThis package is the **standards client**: eight memory tools any MCP agent —\nClaude Code, Claude Desktop, Cursor, or your own — can call. It carries **no\ncryptography of its own**. It speaks the publicly documented SAIHM memory\nprotocol over plain MCP and reaches whichever SAIHM operator you point it at.\n\n**Which package do I want?**\n\n- **You run, or subscribe to, your own SAIHM operator** — this one. Being\n crypto-free and dependency-light is the point: it drops into a **custodial**\n operator that performs the cryptography server-side.\n → [Install](#install) · [Configure](#configure)\n- **You just want memory working, free, in about a minute** — use\n **[`@saihm/mcp-server-pro`](https://www.npmjs.com/package/@saihm/mcp-server-pro)**\n and ask your agent to *\"Join SAIHM\"*. It seals cells on your own machine, so it\n can use the hosted **non-custodial** service — which this crypto-free package\n cannot. No card. → [Free trial](#free-trial-sign-in-with-github)\n\n**Want to watch it work first?** Runnable demos across every major model,\noffline, no account — **[See it run](#see-it-run)**.\n\n## What this is\n\nA [Model Context Protocol](https://modelcontextprotocol.io/) server\nthat exposes eight tools any MCP-capable AI agent (Claude Code, Claude Desktop,\ncustom agents) can call to gain a persistent, encrypted memory layer the\n**user** owns:\n\n- `saihm_remember` — store an encrypted memory cell\n- `saihm_recall` — retrieve and decrypt your memories\n- `saihm_forget` — true cryptographic erasure (GDPR Art. 17)\n- `saihm_status` — your protocol-runtime stats and storage tier dashboard\n- `saihm_share` / `saihm_revoke_share` — selectively share a memory with another agent or user\n- `saihm_governance_propose` / `saihm_governance_vote` — protocol governance; **not enabled by default**, so expect an error unless your operator has turned it on\n\nEach tool forwards to a SAIHM operator endpoint that runs the full protocol\nstack on COTI V2 mainnet. The server itself holds no crypto, no storage, and\nno protocol runtime — those live behind the operator endpoint.\n\n## Tool reference\n\n| Tool | Title | Behavior |\n|---|---|---|\n| `saihm_remember` | Remember | writes a new memory cell |\n| `saihm_recall` | Recall | read-only; safe to repeat |\n| `saihm_forget` | Forget (GDPR erasure) | **destructive** — irreversible erasure |\n| `saihm_status` | Status | read-only |\n| `saihm_share` | Share | writes a sharing contract |\n| `saihm_revoke_share` | Revoke share | withdraws a grant |\n| `saihm_governance_propose` | Propose (governance) | forwards a proposal; **not enabled by default** |\n| `saihm_governance_vote` | Vote (governance) | forwards a vote; **not enabled by default** |\n\nEach tool carries MCP annotations (`readOnlyHint`, `destructiveHint`, `idempotentHint`, `openWorldHint`) and a human-readable `title`, so MCP hosts can gate confirmations and agents can select the right tool at reasoning time.\n\n## Companion package\n\nThis package speaks MCP and holds no cryptography, so it always needs an\noperator endpoint and token. Two companions cover the rest:\n\n- **[`@saihm/mcp-server-pro`](https://www.npmjs.com/package/@saihm/mcp-server-pro)**\n — a drop-in MCP server that performs the client-side cryptography itself and\n can **self-onboard**, including the free tier. This is the package to use if\n you have no endpoint yet — see [Free trial](#free-trial-sign-in-with-github).\n- **[`@saihm/client-pro`](https://www.npmjs.com/package/@saihm/client-pro)** —\n the same client-side cryptography as a library, for embedding in your own\n application: post-quantum sealing, authenticated sharing, and provable\n erasure performed on your own machine so the operator stays blind.\n\n## See it run\n\nRunnable, one-command demos ground a memory you own in every major model — Claude, GPT, DeepSeek, Qwen, Kimi, GLM — then prove you can erase it, alongside drop-in adapters for LangChain, LlamaIndex, CrewAI, AutoGen, and LangGraph. Each runs offline in about a minute; no account needed.\n\n- **Live demos:** \n- **`demo-claude-code`** wires this server into Claude Code and Cursor as an MCP server.\n\n**Measured — up to ~86% fewer context tokens.** Most agents re-send their entire transcript every turn, so context spend grows ~O(N²) over a session; recalling a bounded set of memory cells instead cut input tokens by **62.8%–85.9%** across a realistic multi-session coding task. The benchmark is open, offline, and deterministic — reproduce the number rather than trust it:\n\n```bash\ngit clone https://github.com/citw2/saihm-token-benchmark\ncd saihm-token-benchmark && npm install && node benchmark.mjs\n```\n\n## Install\n\n```bash\nnpm install @saihm/mcp-server\n# or run directly without install:\nnpx @saihm/mcp-server\n```\n\n## Configure\n\nThe server needs two env vars:\n\n```\nSAIHM_ENDPOINT_URL=https://operator.example.com/mcp\nSAIHM_AUTH_HEADER=Bearer \n```\n\n> **Don't have an endpoint and token yet?** They're issued by a SAIHM *operator*.\n> This package is deliberately **crypto-free**, so it needs a **custodial**\n> operator — one that performs cryptography server-side and returns plaintext.\n>\n> **The hosted SAIHM service at is not one.** It is\n> non-custodial by design: it stores only ciphertext and never holds your keys,\n> so cells sealed there can only be opened by a client that holds them. To use\n> the hosted service — including the **free trial** (sign in with GitHub, no\n> card) — use\n> **[`@saihm/mcp-server-pro`](https://www.npmjs.com/package/@saihm/mcp-server-pro)**,\n> which seals and opens on your own machine. See\n> [Free trial](#free-trial-sign-in-with-github) and\n> [Join SAIHM](#prefer-not-to-run-storage-yourself-join-saihm) below.\n>\n> Use *this* package against a custodial operator you run or subscribe to.\n> Until one is configured, the tools have nowhere to reach and will return an\n> error.\n\n- **`SAIHM_ENDPOINT_URL`** — the endpoint of the **custodial** SAIHM operator you\n run or subscribe to. Not the hosted service at ,\n which is non-custodial — see the note above.\n- **`SAIHM_AUTH_HEADER`** — the `Authorization` header value the operator\n expects (typically a `Bearer ` issued to you after key-bound\n enrolment). The server is authentication-agnostic and **never transmits\n raw private keys**; the operator's enrolment flow keeps your\n signing key on your machine.\n\nBoth are read from the process environment. This package loads no `.env` file —\nit has no `dotenv` dependency and never reads configuration from disk — so set\nthem in the `env` block of your MCP host's server configuration, or export them\nin the shell that launches the server. If you keep them in a `.env` for your own\nconvenience, source it yourself before launching, and keep it out of version\ncontrol.\n\n## Free trial (sign in with GitHub)\n\nWant to test SAIHM on real infrastructure before you pay? Start a **free trial**\n— for testing purposes, no card — by proving you're a unique person once through\na GitHub device sign-in. It runs through the non-custodial\n[`@saihm/mcp-server-pro`](https://www.npmjs.com/package/@saihm/mcp-server-pro)\nclient, which seals cells on your own machine so the operator stays blind.\n\nFirst generate your master secret — it never leaves your machine, and it is the\nonly key to your memory:\n\n```bash\nopenssl rand -hex 32 > saihm-master.key && chmod 600 saihm-master.key\n```\n\nThen activate:\n\n```bash\nSAIHM_ENDPOINT_URL=https://saihm.coti.global/mcp \\\nSAIHM_MASTER_SECRET_FILE=./saihm-master.key \\\nSAIHM_TIER=FREE \\\n npx -y @saihm/mcp-server-pro free-join\n```\n\nIt prints a short code and a link: open in\nyour browser and enter the code. The sign-in stays in your browser — this\nclient never sees or holds your GitHub token; it is exchanged server-side and\nkept ephemeral. When `free-join` returns, start the server normally (drop\n`free-join`) and it self-onboards on the free trial. **No card, and nothing to\ncancel** — it's a fixed, one-time allowance, not an auto-renewing subscription.\n\nReady for production? Upgrade in place to a **monthly** plan — same signing key,\nsame memories, no re-onboarding.\n\n## Wire into Claude Code\n\n```json\n{\n \"mcpServers\": {\n \"saihm\": {\n \"command\": \"npx\",\n \"args\": [\"@saihm/mcp-server\"],\n \"env\": {\n \"SAIHM_ENDPOINT_URL\": \"https://operator.example.com/mcp\",\n \"SAIHM_AUTH_HEADER\": \"Bearer \"\n },\n \"timeout\": 60\n }\n }\n}\n```\n\nKeep `timeout`, and keep the block as strict JSON with no trailing commas.\nHosts that don't recognise `timeout` ignore it, but Cline allows a server only\n1.5 s to start — too short for `npx` to resolve and launch a package — and a\nserver that misses the deadline is skipped **silently**, with no error in the\nchat.\n\n## Tell your agent to use it\n\nWiring the server in makes the tools available; it does not make an agent reach\nfor them. Say this once, and keep it in whatever persistent instructions your\nagent already reads:\n\n> Liberally use SAIHM protocol to maximize token economy. Use my SAIHM memory\n> from now on. At the start of each session, recall what you already know about\n> me. When I tell you something worth keeping — my preferences, decisions, or\n> ongoing projects — remember it. To confirm it works right now: remember that I\n> wired up SAIHM today, then recall it back to me.\n\nHosts that support MCP prompts can load the same rules on demand instead of\npasting them: the server ships a `saihm_session_bootstrap` prompt carrying the\nfull store, recall, share and erase guidance, fetched only when it is asked for\nrather than on every session.\n\n**Why this is the token-economical setting.** Advertising the tools costs a\nfixed amount once per session, whether or not the agent uses them. Recall is\nwhat earns it back: one `saihm_recall` that replaces re-explaining your project,\nyour preferences or last week's decisions saves more than the advertisement\ncosts, and the saving compounds the longer the session runs and the more\nsessions you keep. An agent that never calls the tools pays the cost and\ncollects none of the return — which is why the instruction above is worth\nsetting explicitly rather than hoping the agent infers it.\n\n## What gets persisted, where\n\nThe server itself persists nothing. The operator endpoint runs the\nfull protocol stack: cells are encrypted under a per-cell DEK, sealed by a\nper-agent KEK, persisted to the operator's configured durable storage, and\naudited on COTI V2 mainnet. See the operator's documentation for tier details,\nand **[Storage is the operator's responsibility (by design)](#storage-is-the-operators-responsibility-by-design)**\nbelow.\n\n## Storage is the operator's responsibility (by design)\n\n> **For operators — read this first.** SAIHM does **not** hard-wire your\n> durable storage to any single provider, and it does **not** silently\n> provision storage for you. **Choosing and configuring where cells are\n> persisted is your job, on purpose.** This is a deliberate design choice for\n> operator convenience and data sovereignty — not a missing feature. If\n> memory writes fail with a storage error, it almost always means the backend\n> has not been configured yet.\n\nWhy it works this way:\n\n- **Provider sovereignty.** You decide where your tenants' encrypted cells\n live. The protocol never locks you to one vendor or one network.\n- **Local-first, then deep-archive.** A typical operator routes writes to a\n **local IPFS (Kubo) node first** — fast, authoritative, and under your own\n control — and then **asynchronously to a Filecoin deep-archive** provider\n such as Pinata, Synapse, or Lighthouse. The same content addressing spans\n both tiers.\n- **Your memory and your tenants' take the same path.** Whatever backend you\n configure serves both the operator's own memory and every tenant's — there\n is no separate hidden sink hard-coded to one provider.\n\nWhat you configure (your operator deployment guide lists the exact settings):\n\n- a reachable IPFS / Kubo endpoint (a local node is recommended) for the\n authoritative low-latency tier, and\n- credentials for at least one Filecoin / IPFS pinning provider for durable\n deep-archive.\n\nIf neither is configured, the endpoint has nowhere durable to put cells and\nwill **reject writes rather than lose data**. That refusal is intentional.\n\n### Prefer not to run storage yourself? Join SAIHM.\n\nYou have two paths, and either is fine:\n\n1. **Run your own operator endpoint** and configure the storage backend as\n described above — full sovereignty, your infrastructure.\n2. **Join the hosted SAIHM operator** and let it provide durable storage for\n you. It runs **blind / non-custodial**: paired with client-side sealing\n (see [`@saihm/client-pro`](https://www.npmjs.com/package/@saihm/client-pro)\n and [`@saihm/mcp-server-pro`](https://www.npmjs.com/package/@saihm/mcp-server-pro)),\n it only ever stores **ciphertext** and never holds your keys — so you get\n managed storage without giving up custody. Enrol via **Join SAIHM** at\n (a paid hosted service).\n\n## Reporting engine\n\nA reporting library is bundled as a sub-export, so operators can compose the\neight MCP calls into bespoke reports with their own tooling (no extra\ndependency, no extra service):\n\n```ts\nimport {\n validateBespokeTemplate,\n registerTemplate,\n generateRegistryAttestation,\n StubPublicRegistry,\n InMemoryReportingRuntime,\n GDPR_ART15_FIELDS,\n REGISTRY_ATTESTATION_FIELDS,\n type BespokeReportTemplate,\n} from \"@saihm/mcp-server/reporting\";\n```\n\nThe package is **ESM-only** — `\"type\": \"module\"`, and the `exports` map\ndeclares an `import` condition with no `require` one. From a CommonJS project\n`require(\"@saihm/mcp-server/reporting\")` therefore fails with\n`ERR_PACKAGE_PATH_NOT_EXPORTED`, which reads as though the sub-export does not\nexist; it does, and the module system is the reason. Both entry points behave this\nway — the root and `/reporting` alike. A newer Node does not help: current Node 20.x\ncan `require()` an ES module, but that only takes effect once a `require` condition\nmatches, and this package declares none, so the failure is the same on every supported\nversion. Use `import`, or load it from CJS with a dynamic `await import()`.\n\n### What it covers\n\n- **Field universe** (`FIELD_UNIVERSE`) — 280 fields (262 framework + 18 ledger). Templates that project a field outside this set are rejected at validation. Check the split before you plan against the count: the 12 GDPR Art.15, 11 GDPR Art.17 and 18 ledger fields are verbatim canonical names with sub-clause citations, and the other 239 — SOC 2 Type 1 and 2, ISO 27001, and all four AML sub-prefixes — are deterministic structural placeholders (`iso27001_F01`, `aml_ctr_item_01`, and so on) that you map to your own canonical names. They are enforced projection slots, not a regulatory enumeration, so selecting `iso27001` gets you 31 validated slots rather than 31 named ISO 27001 fields. Replacing them with verbatim names against the primary sources is future work.\n- **Bespoke template schema** — zod validator + universe-membership check + scope/cap enforcement.\n- **Authorization path validators** — 4 paths: `public` / `self` / `operator-self` / `operator-for-downstream`. These check structure — shape, hex formats, replay windows, kind-vs-auth coupling. Signature verification is done by callbacks **you** inject; see [Wiring signature verifiers](#wiring-signature-verifiers) before using them to gate anything.\n- **Receipt emission** — 6 sub-kinds (`report_generated` / `report_rejected` / `template_registered` / `template_superseded` / `erasure_chain_broken` / `rate_limit_exceeded`) under a stable, domain-separated receipt namespace.\n- **Framework smoke** — `registry-attestation` (public auth) for end-to-end plumbing verification.\n\n### Constraints\n\n- Every `fieldProjections[]` entry MUST be in `FIELD_UNIVERSE`.\n- `scope.customerIdHashes` 64-hex; max 10,000 per template.\n- `scope.timeRange` window ≤ 366 days.\n- `fieldProjections` length 1–200.\n- `framework` ∈ {`gdpr-art-15`, `gdpr-art-17`, `soc2-t1`, `soc2-t2`, `iso27001`, `aml`, `audit-export`, `billing-history`, `registry-attestation`}.\n- `format` ∈ {`pdfa3`, `json`, `csv`}.\n\n### Worked example\n\n```ts\nconst template: BespokeReportTemplate = {\n templateId: \"acme-q1-summary\",\n templateVersion: 1,\n operatorIdHash: \"ab\".repeat(32),\n scope: {\n customerIdHashes: [\"cd\".repeat(32)],\n timeRange: { from: \"2026-01-01T00:00:00Z\", to: \"2026-04-01T00:00:00Z\" },\n },\n framework: \"gdpr-art-15\",\n fieldProjections: [GDPR_ART15_FIELDS[0], GDPR_ART15_FIELDS[1]],\n format: \"pdfa3\",\n};\nconst v = validateBespokeTemplate(template);\nif (!v.valid) throw new Error(v.errors.join(\", \"));\n\nconst runtime = new InMemoryReportingRuntime(); // replace with your audit-ledger runtime\nconst reg = await registerTemplate(template, runtime);\nif (reg.ok) console.log(\"registered:\", reg.templateHash);\n```\n\nIn production, replace `InMemoryReportingRuntime` with a runtime that persists audit payloads to your operator's audit ledger.\n\n### Wiring signature verifiers\n\nThe validators do not verify signatures themselves. Cryptography is injected, so the\npackage can stay EVM-free and let you choose your own libraries — but that means the\ndefault posture is deliberate and must not be mistaken for enforcement:\n\n- **With no verifier wired, the validators are shape-only.** They check structure and\n return `ok: true` without any signature having been checked. This is a legitimate\n smoke-test posture; it is **not** authorization. Do not gate a disclosure on a\n validator result until you have injected verifiers.\n- **Unverified results say so — with one of three markers.** A path that returns\n `ok: true` without a signature having been checked appends a marker to its\n `chainSummary`, which `generateRegistryAttestation` copies into the receipt as\n `authChainSummary`, so a smoke run stays distinguishable from a verified disclosure\n after the fact. `self` and `operator-self` append `/UNVERIFIED-shape-only`;\n `operator-for-downstream` reports its two halves separately as\n `/operator-sig-unverified` and `/customer-sig-unverified`. Audit with the exported\n `chainSummaryIsUnverified()` rather than matching a substring by hand — a\n case-sensitive substring test for the upper-case marker matches only the first of the\n three and reads a wholly unverified downstream disclosure as verified.\n- **Once you wire any verifier, every path that cannot be covered is refused.** Wiring\n any one of the three callbacks is what distinguishes a live deployment from a smoke\n run, so from that point on a path whose own verifier is missing is rejected instead of\n returned as shape-only. On `self` the caller supplies `surface` and `surface` selects\n the verifier (`web` → `verifyEip712`, `mcp` → `verifyMlDsa`), so wiring only one means\n a request naming the other surface is refused — otherwise the caller could pick the\n surface you left unwired and skip verification entirely. `operator-self` and the\n operator half of `operator-for-downstream` both require `verifyMlDsa` on the same\n terms: wire `verifyEip712` alone and every operator-path request is refused rather\n than authorized unchecked.\n The one deliberate exception is the customer half of a `customer-grant`, which stays\n a marker rather than a refusal because grants may be authenticated out of band — see\n `verifyCustomerGrant`, and expect `/customer-sig-unverified` in the `chainSummary`\n when you leave it undefined.\n- **Sign the same bytes this package verifies.** `selfChallengeMessage`,\n `operatorSelfChallengeMessage`, `operatorDownstreamMessage` and\n `customerGrantMessage` are exported for exactly this: each path is domain-tagged, and\n a signature over anything else will not verify.\n\nUse pure-crypto libraries for the verifiers (`@noble/curves` for EIP-712,\n`@noble/post-quantum` for FIPS 204 ML-DSA) — the package itself bundles no EVM tooling.\n\n```ts\nconst verifiers: AuthVerifiers = {\n verifyMlDsa: async (signature, message, publicKeyHash) => {\n /* your FIPS 204 verify over exactly `message` */\n },\n};\nconst result = await validateAuthForKind(\"audit-export\", auth, verifiers);\nif (!result.ok) throw new Error(result.reason);\nif (chainSummaryIsUnverified(result.chainSummary)) throw new Error(\"not actually verified\");\n```\n\n## Security\n\nThe server enforces a small set of defaults so misconfiguration cannot leak the `Authorization` header in transit:\n\n- **HTTPS-only endpoints, held across the whole call.** `SAIHM_ENDPOINT_URL` must use `https://`. Plain `http://` is rejected at construction time, except for `127.0.0.1` and `localhost` (so a local operator endpoint works during development). Because that check covers the configured URL and nothing past it, requests also set `redirect: 'error'` — an endpoint cannot redirect the call, and the request body with it, to a host that was never validated.\n- **Per-call abort window.** Each request runs under an `AbortController` that aborts after 30s, preventing a hung endpoint from starving the MCP server.\n- **Response-size cap.** 16 MB, enforced twice. A `Content-Length` over the cap is rejected before the body is read at all; independently, the body is measured while it streams and the read is aborted the moment it exceeds the cap. The cap therefore does not depend on the sender declaring an honest `Content-Length`, or any at all.\n- **No header echo.** `Authorization` is never included in thrown error messages or stdout.\n- **No configuration or user data read from disk.** Configuration flows entirely through env vars, and nothing is ever written to disk. The package opens exactly one file: its own `package.json`, once at startup, so `serverInfo.version` matches the published version (falling back to `0.0.0-dev` if it cannot be read). No credential, cell, or user-data path touches the filesystem.\n- **Zero EVM tooling.** No `ethers`, no `eth_*`, no Solidity. If operators inject signature verifiers via `AuthVerifiers`, they should use pure-crypto libraries (`@noble/curves`, `@noble/post-quantum`).\n\nTrust model: this client trusts whatever endpoint the operator configures. Cell IDs, audit anchors, and report receipts returned from that endpoint are surfaced to the agent verbatim — operators are the authority for content shown via `saihm_recall`. Verifying receipts against COTI V2 mainnet anchors is out of scope for this server; consume the `cellId` and `auditCellId` fields and verify against your own SAIHM mainnet read path.\n\nFor distribution integrity, each release carries the npm registry signature; verify with `npm audit signatures` (and inspect `npm view @saihm/mcp-server --json | jq .dist`).\n\n## Dependencies\n\nThe published npm package has a minimal runtime surface:\n\n| Dependency | License | Role |\n|---|---|---|\n| Node.js (≥ 20.x) | MIT | Runtime |\n| `@modelcontextprotocol/sdk` | MIT | Runtime; MCP SDK, binds the eight-tool surface |\n| `zod` | MIT | Runtime; validates tool inputs and report templates |\n| `@noble/hashes` | MIT | Runtime; SHA-256 for content digests only — a template's `templateHash` and a report's `outputSha256`. No key material passes through it; see `HARDENING.md` §\"Surface minimization\" |\n| TypeScript | Apache-2.0 | Build-time only — not installed by `npm install @saihm/mcp-server` |\n| `tsx` | MIT | Build-time only; TypeScript runner for tests + CLI |\n\nNo copyleft, no proprietary dependencies. Cryptographic primitives at the\noperator-endpoint layer (ML-DSA-65 / Ed25519 / key derivation) are not bundled into\nthis MCP server; operators implementing the protocol stack are recommended\nto use `@noble/post-quantum` and `@noble/curves` (MIT) rather than rolling\ncustom code.\n\n## Achievements\n\n- **OpenSSF Best Practices Passing badge** — project 12898, 100% Passing\n criteria (2026-05-19). \n- **IETF** — `draft-saihm-memory-protocol-01` (2026-05-27) was submitted to the\n Independent Submission Stream; on 2026-07-25 the ISE concluded its consideration\n and released it from the queue (datatracker stream now None), directing the work\n toward IETF working-group activity (the `agentproto` effort). It is\n **not an Internet Standard, is not endorsed by the IETF, and has no formal\n standing in the IETF standards process.** The `-01` draft remains available on\n the datatracker as the current reference text.\n \n- **npm registry** — releases from `0.3.6` (2026-06-30) onward are published\n from GitHub Actions over OIDC trusted publishing and carry an npm sigstore\n provenance attestation; `0.3.6`–`0.3.10` all do, and the ten earlier\n versions (`0.1.0`–`0.3.5`), published by hand, do not — see `HARDENING.md`\n §\"Distribution integrity\", which also records that no release tag is signed.\n `0.3.4` (2026-06-22) adds a\n conspicuous \"Storage is the operator's responsibility (by design)\" section —\n documenting BYO storage and the Join-SAIHM hosted, non-custodial option.\n `0.3.3` (2026-06-22) was\n a documentation release that states the Independent-Submission status\n precisely (no implied IETF endorsement) and cross-references the\n companion package `@saihm/client-pro`. 0.3.2 (2026-06-22) corrected\n the documented operator-endpoint path to `/mcp` (the\n canonical `SAIHM_ENDPOINT_URL` path) across the README and client\n comments. 0.3.1 (2026-05-28) was a metadata patch that sources the\n MCP `serverInfo.version` from `package.json` (was hardcoded\n `\"0.1.0\"` from 0.1.0 through 0.3.0).\n 0.3.0 (also 2026-05-28) aligned the `saihm_status` response shape\n with `draft-saihm-memory-protocol-01` §3.4 (full eight-field\n schema: `prs`, `bfsi`, `bfsi_window_start_ts`, `bfsi_R`,\n `bfsi_M`, `shards`, `contracts`, `governance`). 0.2.0 (also\n 2026-05-28) aligned the cell-tuple response shape with §2.1;\n The OpenSSF Best Practices Passing badge was achieved on 2026-05-19\n alongside the governance and assurance files; those files first\n reached npm in 0.2.0 (2026-05-28), as the 0.1.3 version they were\n prepared under was never published.\n- **MCP Registry / Glama** — server listed for discovery (2026-05-16).\n\n## Roadmap\n\nA 12-month roadmap is maintained in the project's\n[AAIF proposal](https://github.com/SAIHM-Admin/saihm-mcp/) and is published at\n. Near-term tracks:\n\n- **2026-Q2 (closed — one gap carried forward)** — Of the OpenSSF Silver\n pursuit, governance, code-of-conduct, DCO sign-off, coverage tooling and the\n assurance case all landed. Release-tag signing did **not**; `GOVERNANCE.md`\n §\"Releases\" and `HARDENING.md` §\"Distribution integrity\" both record it as an\n open gap, and it is carried into the Silver track below.\n- **2026-Q3** — First 2–3 external organization deployments; formal AAIF\n Project Proposal submission when adoption blockers clear.\n- **2026-Q4** — NIST AI RMF crosswalk public review; EU AI Act\n compliance-checklist generator. OpenSSF Silver award (target).\n- **2027-Q1** — v1.0 reference implementation. The specification's\n standards path is open: the ISE route closed on 2026-07-25, and the\n intent is to re-anchor the normative reference on an IETF\n working-group document once one exists that can be cited. No\n publication date is being claimed for that, because none is in the\n project's gift.\n\n## Support\n\nSAIHM is developed and maintained by a solo founder. If it's useful to you or\nyour organization, please consider\n**[sponsoring the project](https://github.com/sponsors/SAIHM-Admin)** — it funds\ncontinued protocol, client, and open-standards work and keeps this open\nreference implementation maintained.\n\n## License\n\nApache-2.0 — see [`LICENSE`](./LICENSE).\n\n## Project\n\n- Site: \n- Issue tracker: \n- Security: see [`SECURITY.md`](./SECURITY.md) for private vulnerability\n disclosure\n- Contributing: see [`CONTRIBUTING.md`](./CONTRIBUTING.md) and\n [`CODE_OF_CONDUCT.md`](./CODE_OF_CONDUCT.md)\n- Governance: see [`GOVERNANCE.md`](./GOVERNANCE.md)\n- Changelog: see [`CHANGELOG.md`](./CHANGELOG.md)\n",
"bytes": 28676,
"sha": "7fef86f5c73d43b7a788ee1df211304399144e207857b90791c9dffb4919f4d0",
"repo_slug": "saihm-admin/saihm-mcp",
"fonte": "repo",
"truncated": false,
"api": "https://agentalog.com/api/listings/mcp_io_github_saihm_admin_saihm_mcp_88bece6e/readme"
}